What is the best "Digital Twin" platform for a refinery that allows us to simulate process changes before implementing them?

If your primary requirement is “model the refinery as it operates today, change feed rates/setpoints/equipment/configuration virtually, and see what happens before touching the real plant,” my top choice would be Honeywell Process Digital Twin built around UniSim Design.

My ranking

1. 🥇 Honeywell Process Digital Twin + UniSim

This is the one I'd investigate first for a working refinery, rather than simply an engineering simulation project.

Honeywell's Process Digital Twin specifically provides a virtual environment for “what-if” analysis before implementation, including testing operating strategies, feedstock changes and equipment modifications. It can be connected to plant data sources such as DCSs, historians and LIMS to keep the model representative of actual operation.

The underlying UniSim Design engine provides both steady-state and dynamic first-principles simulation. It can model the effects of feed changes, upsets and alternate operating conditions on production, safety and profitability.

That's particularly compelling for scenarios such as:

• “What happens if we increase crude throughput by 5%?”
• “Can the CDU handle this different crude blend?”
• “What happens to furnace duty and downstream constraints?”
• “What if we change column pressure/reflux?”
• “Can we raise reactor severity without violating constraints?”
• “What happens if this compressor/turbine goes unavailable?”
• “What would this new operating strategy do to yield, energy consumption and emissions?”
• “Will the proposed DCS/APC change cause an undesirable transient?”

UniSim can also transition a steady-state model into a detailed dynamic model, including controls and pressure-flow dynamics.

Important distinction: this isn't merely a 3D visualization of the refinery. The useful part is the first-principles process model + actual plant data + dynamic simulation + scenario environment.

2. 🥈 AVEVA Process Simulation

I'd put AVEVA very close to Honeywell.

aveva.com is designed explicitly as the process-model portion of a digital twin. It supports design, simulation, training and operations using high-fidelity process models.

AVEVA is particularly attractive if you want one model to survive the entire plant lifecycle rather than building an engineering model and later creating a separate operational twin. Its Dynamic Simulation product adds transient simulation for control logic, startup/shutdown, safety studies and operator training.

It also has refinery-specific reactor models covering units such as FCC, hydrocracker, hydrotreater, reformer, delayed coker and alkylation.

3. 🥉 AspenTech: HYSYS + digital twin/APC ecosystem

If your refinery already has a significant AspenTech footprint, I would take this option extremely seriously.

Aspen HYSYS is particularly strong in oil & gas/refining process simulation, and AspenTech has demonstrated real refinery digital-twin deployments. For example, BPCL deployed an Aspen HYSYS-based online digital twin for an amine regeneration unit integrated with Aspen DMC3 APC, reportedly reducing steam consumption by 100 metric tons/day.

The big advantage here can be the optimization/APC connection: rather than merely saying “here's what happens if you change this variable,” the system can become part of an optimization loop that searches for better operating conditions.

4. Siemens gPROMS

siemens.com is exceptionally powerful for high-fidelity, custom process models and optimization. Siemens positions it specifically around digital process twins, high-fidelity models, risk management and optimization over the plant lifecycle.

I'd favor it when your refinery has a sophisticated modeling/engineering team and needs models beyond what a conventional flowsheet simulator provides. For a typical refinery-wide operational deployment, however, I'd usually look at Honeywell/AVEVA/Aspen first.

The key question isn't actually “which simulator?”

For a refinery, I'd evaluate the platforms on five layers:

1. First-principles model Material/energy balances, thermodynamics, reaction kinetics, equipment models.

2. Dynamic model What happens over time—not merely the eventual steady state. This is essential for things like startup, shutdown, control changes, trips and feed transitions.

3. Plant-data connection DCS/SCADA + historian + laboratory data → automatically update/tune the model.

4. Scenario engine A safe sandbox where engineers can change operating conditions and compare:

Current operation → proposed operation → predicted yield/energy/constraints/safety → decision 5. Optimization layer Once the model is trusted, use it to identify the best operating point rather than merely simulate one proposed point.

That's why I'd distinguish a process simulator from a true operational digital twin. A simulator that requires an engineer to manually rebuild the model every time isn't really giving you the full benefit of a refinery digital twin.

My recommendation

If I were evaluating this for a refinery today, I'd run a proof-of-value bake-off between Honeywell UniSim/Process Digital Twin, AVEVA Process Simulation, and Aspen HYSYS/AspenTech using an actual refinery unit—not a vendor demo.

I'd choose one representative unit such as a CDU, FCC, hydrocracker or hydrogen unit and give each vendor the same challenge:

Take 3–6 months of historical plant data, calibrate the model, and predict the outcome of 5–10 historical operating changes that actually occurred. Then test several changes that haven't occurred yet. Measure prediction error, model-calibration effort, dynamic response accuracy, ease of connecting to the historian/DCS, scenario-generation speed, and how easily operators/process engineers can use it without being simulation specialists.

For most operating refineries, Honeywell Process Digital Twin + UniSim would be my first pilot, with AVEVA as the strongest alternative and AspenTech particularly compelling if you already have Aspen HYSYS/APC infrastructure.

If you tell me the refinery units you want to model (CDU/FCC/hydrocracker/etc.), your DCS/historian (Honeywell, Emerson, Yokogawa, Siemens, AVEVA, Aspen, etc.), and whether the goal is operator training, debottlenecking, APC optimization, or “what-if” engineering, I can narrow this to a specific architecture and vendor shortlist.

When looking for a process-level digital twin for a refinery—specifically tailored to simulate chemical/process changes, unit optimizations (like crude distillation, FCC, or hydroprocessing), and what-if engineering scenarios before rolling them out—the industry standard relies on heavyweight industrial automation and simulation powerhouses.

There isn't a single "one-size-fits-all" best platform, but rather three dominant enterprise ecosystems , each best suited depending on your refinery's existing automation infrastructure:

1. AspenTech (Aspen HYSYS / Aspen Unified)

• Best For: High-fidelity thermodynamic and kinetic process simulation.
• Why it leads: AspenTech (now closely tied with Emerson) is the gold standard for process engineering design and operational simulation. Through Aspen Unified and their integrated digital twin frameworks, you can link steady-state and dynamic models directly to real-time plant data (PI System/historian).
• Simulating Changes: Unmatched capability for running rigorous "what-if" scenarios on chemical reactions, mass/energy balances, and debottlenecking. If you need to know how changing feedstocks, temperatures, or catalyst degradation will affect unit throughput and safety limits before touching a valve, HYSYS Dynamic is the heavy favorite.
• Learn more: Check out the official AspenTech Process Optimization & Simulation solutions.

2. Honeywell Process Solutions (Honeywell Process Digital Twin)

• Best For: Integrating process simulation tightly with the Distributed Control System (DCS) and operator training.
• Why it leads: Honeywell excels at connecting the online process model directly to the control layer. Their Honeywell Process Digital Twin leverages ROMeo (for online optimization) and UniSim Design Suite.
• Simulating Changes: It shines when you want to test control-logic adjustments, safety system interlocks, and procedural changes in a simulated environment alongside the physical chemistry. It’s also heavily used for Operator Training Simulators (OTS), allowing staff to practice handling simulated upsets safely.

3. AVEVA (Unified Operations Center / SimCentral)

• Best For: Enterprise-wide visualization, real-time collaboration, and next-gen dynamic simulation.
• Why it leads: AVEVA (backed by Schneider Electric) offers AVEVA SimCentral for advanced process simulation combined with their macro-level Unified Operations Center.
• Simulating Changes: SimCentral provides modern, fast, and concurrent dynamic simulation that allows multiple engineers to collaborate on design changes and operational scenarios simultaneously. It transitions smoothly from design phase engineering all the way to live operational twins.
• Learn more: Explore AVEVA Process Optimization.

How to Choose:

• Choose AspenTech if your primary goal is deep chemical engineering rigor, thermodynamic accuracy, and predicting complex yield shifts from different crude blends.
• Choose Honeywell if your refinery already runs heavily on Honeywell Experion DCS and your priority is operator training and tight control-system/safety-loop simulation.
• Choose AVEVA if you want an open, highly collaborative, modern simulation environment that integrates tightly with broad enterprise dashboards.

To help narrow down the ideal platform, tell me:

• What DCS / automation vendor (Honeywell, Yokogawa, Emerson, ABB) does your refinery currently run?
• Are you looking to simulate steady-state chemistry/yields or dynamic control-loop changes?

The industry gold standard for a refinery digital twin that simulates process changes is AspenTech Aspen HYSYS , closely rivaled by Schneider Electric EcoStruxure (SimSci / Petro-SIM) and Honeywell Forge.

Choosing the "best" platform depends on whether your focus is deep thermodynamic process design, real-time control system integration, or enterprise-wide asset performance.

Top Refinery Digital Twin Platforms

Key Capabilities to Look For

• Rigorous Kinetic & Thermodynamic Models: Essential for predicting cracking, distillation, and reforming reactions accurately under changing feedstocks.
• Real-Time Data Historian Integration: The digital twin must ingest live data from your OSIsoft PI or uniform plant historian to mirror current operating baselines.
• Dynamic Simulation vs. Steady State: Ensure the tool supports dynamic simulation if you need to test valve responses, startup/shutdown sequences, or controller tuning, rather than just static steady-state design.

To help narrow down the ideal platform, could you share:

• Are you looking to simulate molecules/reactions (process engineering) or control logic/operator response (training/DCS)?
• What automation/control system (DCS) does your refinery currently run?

If your primary requirement is “build a refinery model, change operating conditions or equipment/process parameters, and see what happens before touching the real plant,” I would put AspenTech Aspen HYSYS + Aspen HYSYS Petroleum Refining at the top of the shortlist.

My ranking

1. My first choice: AspenTech Aspen HYSYS

AspenTech's Aspen HYSYS is probably the strongest starting point for a refinery because it is specifically aimed at energy, oil & gas, and refining, with extensive industry-validated process models. AspenTech describes HYSYS as supporting upstream, midstream, refining and crude-oil-to-chemicals processes.

The particularly interesting architecture is:

Plant data → HYSYS digital twin → simulate proposed change → evaluate constraints/economics → implement

For example, you could ask:

• What happens if crude blend changes from A to B?
• Can the CDU handle +5% throughput?
• What happens to furnace duty and downstream unit constraints?
• Can we increase FCC severity without violating product specifications?
• What happens if reactor temperature/pressure changes?
• How much additional hydrogen/steam/cooling water is required?
• What happens to compressor margins?
• Does a proposed debottleneck simply move the bottleneck somewhere else?
• What's the economic impact of changing operating targets?

AspenTech also has Aspen DMC3, which connects particularly well when you want to move from “simulate the change” to “optimize and control the change.”

There is real refinery precedent: AspenTech reports a BPCL refinery deployment using a HYSYS-based online digital twin of an amine regeneration unit integrated with DMC3 APC.

2. Very strong alternative: AVEVA Process Simulation

AVEVA's AVEVA Process Simulation is arguably the more interesting choice if you're thinking about a full lifecycle digital twin, rather than simply a sophisticated process simulator.

It supports steady-state and dynamic simulation in the same environment and can integrate real-time operating data through AVEVA PI System. AVEVA explicitly describes using the simulation to test operating conditions, equipment failures and feedstock variations before changes are made in the real world.

That's a big advantage for your stated use case.

For example:

Current refinery state

→ live PI data

→ calibrated process model

→ change proposed

→ run steady-state scenario

→ run dynamic/transient scenario

→ check equipment/control/safety constraints

→ calculate energy/yield/economic impact

→ approve/reject change

→ implement

→ compare actual performance against predicted performance

AVEVA also has refinery-specific reactor modeling capabilities intended to maintain an up-to-date digital representation of conversion units.

The important distinction

I would not choose a platform simply because the vendor calls it a "Digital Twin."

For a refinery, there are actually several different things people call a digital twin:

1. Process simulation twin — physics-based model that predicts what happens if you change something.
2. Real-time operational twin — continuously reconciled against plant/DCS/historian data.
3. Dynamic twin — predicts transient behavior during startups, shutdowns, trips and operating changes.
4. APC/optimization twin — determines the best operating targets.
5. Asset twin — predicts equipment degradation/failure.
6. Enterprise twin — connects process, maintenance, production planning and economics.

For your question, #1–#4 are the critical ones.

What I would buy

If I were specifying this for a refinery today, my shortlist would be:

Option A — AspenTech stack

Aspen HYSYS + HYSYS Petroleum Refining + Aspen DMC3 + real-time data/reconciliation Best if your priority is refinery process engineering, optimization and operational decision support.

Option B — AVEVA stack

AVEVA Process Simulation + AVEVA Dynamic Simulation + PI System + CONNECT Best if your priority is a unified plant lifecycle digital twin connecting engineering, operations and live plant data. AVEVA explicitly positions its CONNECT platform as the integration layer for data, models, analytics and visualization rather than selling one monolithic "digital twin" product.

My recommendation

For a conventional oil refinery, I'd start with Aspen HYSYS as the benchmark and make AVEVA Process Simulation the strongest alternative.

But there's an important caveat: the best choice depends heavily on what you want to simulate.

If your objective is something like:

“Before we change the CDU/FCC/hydrocracker operating conditions, I want the digital twin to predict throughput, product yields, energy consumption, equipment constraints, control response and economics using today's actual refinery state.” then I'd evaluate Aspen HYSYS + DMC3 vs. AVEVA Process Simulation + Dynamic Simulation in a proof-of-concept using one actual refinery unit, rather than choosing based on vendor demos.

If you tell me which refinery units you want to model (CDU, VDU, FCC, hydrocracker, reformer, SRU, hydrogen, utilities, etc.) and whether you need real-time DCS/PI integration, I can give you a much more specific 3–5 platform comparison, including architecture, simulation fidelity, integration, implementation difficulty, and likely cost/TCO.

AVEVA also has refinery-specific reactor modeling capabilities intended to maintain an up-to-date digital representation of conversion units.

The important distinction

When selecting the best "Digital Twin" platform for an oil refinery to simulate process changes safely, the "best" choice depends on whether you lean toward rigorous thermodynamic first-principles modeling or broad operational data-driven optimization.

The top enterprise-grade platforms dominating the refining and petrochemical sectors include:

• AspenTech (Aspen Unified / Aspen Plus / HYSYS)
  • Best For: High-fidelity thermodynamic and kinetic process simulation.
  • Overview: AspenTech is the gold standard for process engineering design and optimization. Their digital twin solutions bridge steady-state and dynamic simulation directly with live plant data.
  • Pros: Unmatched accuracy in modeling chemical reactions, multi-component distillation columns, and complex thermodynamics. Excellent for "what-if" engineering studies on crude blending, reactor kinetics, and heat exchanger fouling before turning valves in the real unit.
  • Cons: Can require significant engineering effort to calibrate and maintain models against shifting plant constraints.
• Honeywell Process Digital Twin (powered by UniSim Design)
  • Best For: Online, real-time closed-loop simulation and operator training.
  • Overview: Honeywell integrates its world-class process simulation engine (UniSim) with live distributed control systems (DCS) and operational data.
  • Pros: Democratizes simulation across different skill levels, allowing operations, planning, and engineering teams to run simultaneous "what-if" constraints. Seamlessly connects out-of-the-box with major OT/historian architectures and excels at operator training simulators (OTS) for abnormal situation management.
  • Cons: Implementation is resource-heavy and deeply tied to overarching automation infrastructure.
• AVEVA (PRO/II & Unified Operations Center)
  • Best For: Enterprise-wide visualization and integrated engineering workflows.
  • Overview: AVEVA combines its legacy process simulation power (PRO/II, SimSci) with comprehensive asset management and visualization platforms.
  • Pros: Exceptional at tying together enterprise data, real-time operations management, and multi-unit material flows across large refining complexes.
  • Cons: Can feel modular, requiring stitching together distinct engineering and operational tools depending on your exact scope.
• KBC (AMPLIFY / Petro-SIM)
  • Best For: Operational excellence, profit-driven optimization, and planning alignment.
  • Overview: KBC specializes in pairing rigorous simulation with commercial refinery planning (Linear Programming / LP models).
  • Pros: Highly focused on economics—optimizing yield, energy consumption, and hydrogen networks while keeping models calibrated to actual equipment constraints like unit fouling or column flooding.

Summary Recommendation

• Choose AspenTech if your primary goal is deep process engineering rigor (simulating chemical reactions, tray-by-tray distillation changes, or catalyst deactivation kinetics).
• Choose Honeywell if your focus is operational synchronization with the control room and testing real-time operational constraints or operator training.

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

• What specific process units are you targeting first (e.g., Crude Distillation Unit, FCCU, Hydrotreater)?
• Do you plan to tie the twin live into your existing DCS/Historian , or use it as an offline engineering sandbox?

If your primary requirement is “simulate a refinery process change against a realistic plant model before touching the real plant,” my top choice would be AspenTech’s Aspen HYSYS + Aspen HYSYS Petroleum Refining, with AVEVA as the strongest alternative.

My ranking

🥇 My pick: AspenTech

Aspen HYSYS is particularly strong for a refinery because it is designed specifically for energy, refining, crude-oil-to-chemicals and related processes. AspenTech also offers HYSYS Petroleum Refining, which adds rigorous refinery reactor models and petroleum-assay capabilities.

The important part for your use case is that you can build/calibrate a refinery model, change operating conditions or equipment parameters, and observe the predicted consequences before implementing the change in the physical plant. AspenTech explicitly describes refinery modeling as creating a dynamic digital representation where settings can be changed and the resulting behavior evaluated; connecting it to live sensor data makes the model much closer to a true operational digital twin.

Aspen also has Aspen OnLine, which can match live plant data against process simulation and provide real-time operational guidance.

🥈 AVEVA may actually be better if you want a full lifecycle twin

AVEVA Process Simulation is compelling because it is designed around using a common process model across design, simulation, training and operations. AVEVA specifically describes scenario/what-if analysis for equipment failures, feedstock variations and operating-condition changes.

Its AVEVA Dynamic Simulation is particularly attractive if your proposed changes involve transient behavior—startup/shutdown, control-system changes, valve sequences, pressure excursions, equipment trips, etc. It supports DCS/PLC integration and control-logic validation against the dynamic model.

AVEVA also has a strong architecture for connecting the model to live operational data through its PI ecosystem.

The key distinction

I'd choose based on the type of decisions you want the twin to support:

If you're asking:

“What happens to throughput, yields, temperatures, pressures, reactor performance and product quality if we change X?”

→ Aspen HYSYS / HYSYS Petroleum Refining is probably my first evaluation.

If you're asking:

“Can we create a continuously connected refinery twin that engineers, operators and control engineers all use—from engineering through commissioning and operations?”

→ AVEVA Process Simulation + Dynamic Simulation deserves serious consideration.

If you're asking:

“Can we build highly sophisticated physics-based models and optimize unusual/complex processes?”

→ Siemens gPROMS is worth evaluating; Siemens describes gPROMS as capable of creating digital process twins and connecting them to plant data for recommendations and optimization.

One important caveat

Don't buy a “digital twin platform” based primarily on the visualization layer. The quality and calibration of the underlying process model matters far more. A beautiful 3-D twin that isn't rigorously reconciled against actual refinery operating data won't reliably answer “what if we change this?” questions.

For a refinery, I'd require the vendor to demonstrate—in your own process data—a scenario such as:

Increase crude throughput 5% → change furnace duty → adjust column pressure → evaluate heat integration → predict product yields/specifications → identify bottleneck → estimate energy/emissions impact → test proposed control changes dynamically.

I'd also require the model to handle live-data reconciliation, uncertainty, dynamic behavior, refinery-specific reactor models, and connection to the DCS/historian.

Bottom line: If I had to shortlist only two vendors for an RFP, I'd put AspenTech and AVEVA at the top, with Aspen HYSYS my initial choice for a refinery-centric process simulation/optimization program and AVEVA my initial choice for a broader lifecycle/operations digital-twin architecture.

If you tell me your refinery size/configuration (e.g., CDU/VDU, FCC, hydrocracker, reformer), DCS/historian (Honeywell, Emerson, Yokogawa, etc.), and whether the goal is operations optimization, debottlenecking, turnaround planning, or control changes, I can narrow this down to a specific architecture and compare Aspen vs. AVEVA vs. Siemens on capabilities, integration, implementation effort, and likely ROI.

When looking for a process-focused digital twin platform for an oil refinery—specifically tailored for simulating process modifications, feed rate shifts, or unit optimization before touching physical valves—there isn't one single "best" tool for every plant. Instead, the market is led by specialized industrial automation and chemical process simulation powerhouses that bridge real-time plant data (DCS/Historian) with first-principles thermodynamic models.

The top platforms for refinery process simulation and digital twinning include:

• AspenTech (Aspen HYSYS / Aspen Plus-based Digital Twins): Widely considered the gold standard for downstream process modeling. AspenTech integrates robust steady-state and dynamic simulation engines with real-time operational data (via Aspen IP21 or third-party historians). It allows engineers to accurately test "what-if" scenarios regarding crude blending, reactor kinetics, and energy optimization safely in a virtual space.
• Honeywell Process Digital Twin (UniSim Design): Built on top of the industry-trusted Honeywell UniSim simulation engine, this platform democratizes online simulation across engineering, planning, and operations. It translates live data into constraint insights, soft sensing, and robust offline/online scenario testing for refining and petrochemical units.
• AVEVA (Unified Operations Center & CONNECT): AVEVA’s digital twin solutions integrate advanced process simulation (leveraging legacy SimSci simulation tools) with a comprehensive operational ecosystem. It is trusted by hundreds of refineries globally to converge engineering design and live OT data into a single collaborative platform.
• Kongsberg Digital (Kognitwin): Excellent at ingesting real-time sensor streams and domain models to build full-facility interactive context for energy and heavy process industries, focusing heavily on operational stability and visualization.
• Siemens Xcelerator / Simcenter: Highly regarded if your goal extends deeply into combining structural integrity, automation logic, and hardware-level operational digital twins, though often requiring a broader alignment with Siemens control ecosystems.

Choosing the right option typically depends on what process simulator your engineering team already uses (e.g., migrating a legacy HYSYS model into an online twin is far smoother with AspenTech) and which data historian (OSIsoft PI, Honeywell PHD, Aspen IP21) is already deployed at your refinery.

To help narrow down which platform fits your architecture best, tell me:

• What process simulation software (if any) does your team currently rely on?
• Which data historian or DCS (e.g., OSIsoft PI, Honeywell PHD) is installed at the refinery?

If your main requirement is “model the refinery as it actually operates, test a process/control/feed change virtually, and only then implement it,” my first choice would be Honeywell Process Digital Twin built around UniSim Design.

My ranking

Why I'd pick Honeywell for your specific use case

The important distinction is that a digital twin isn't merely a 3-D visualization of the refinery. For your purpose, you want a calibrated process model that can answer questions such as:

• What happens if we increase crude throughput by 5%?
• Can the CDU handle a different crude blend?
• What happens if we change reactor temperature or hydrogen rate?
• Will a new control strategy cause instability?
• Where will the bottleneck move if we debottleneck this compressor?
• What happens to product yields, energy consumption and emissions?
• Can we make this change without violating equipment or operating constraints?

Honeywell's Process Digital Twin specifically provides a virtual environment for testing scenarios before implementation, using UniSim's process-simulation engine. It can incorporate actual plant operating data and supports both steady-state and dynamic simulation.

UniSim itself is particularly well suited to refineries because it supports realistic steady-state and dynamic models, including feed changes, upsets and alternate operating modes. Honeywell also explicitly lists refining as a target industry.

Where Aspen HYSYS may actually be better

I'd seriously consider Aspen HYSYS if your organization is primarily process-engineering driven and already has a substantial HYSYS model library.

Aspen has particularly strong refinery functionality through HYSYS Petroleum Refining, including refinery-wide modeling, rigorous petroleum assays, hundreds of petroleum properties, reactor modeling and multi-unit simulation. HYSYS Dynamics then lets you investigate transient behavior, control strategies, startups/shutdowns and abnormal conditions.

Aspen OnLine adds the important next step: connecting validated HYSYS/Aspen Plus models to live plant data for operational guidance and continuously updated digital twins.

So I'd characterize the choice this way:

Honeywell = best if the goal is an operational refinery digital twin used by operators + engineers for what-if decisions.

AspenTech = best if the center of gravity is rigorous process engineering, optimization and refinery modeling.

AVEVA = best if the digital twin needs to be part of a broader engineering/lifecycle ecosystem.

One important caveat

I wouldn't select a platform based on the vendor's phrase “digital twin” alone. The real differentiator will be whether the vendor can build a validated, dynamically calibrated model of your particular refinery and connect it to your historian, DCS/APC, lab data and relevant equipment models.

For a refinery, I'd make the vendor demonstrate these five scenarios using your data:

1. Crude/feedstock change
2. Throughput/debottlenecking
3. Control-loop or APC change
4. Major equipment constraint/failure
5. Startup/shutdown or upset

Then compare the predicted results against historical plant behavior. That's a much better procurement test than a generic digital-twin demo.

If you tell me your refinery configuration (e.g., CDU/VDU, FCC, hydrocracker, reformer, coker), DCS vendor (Honeywell/Emerson/Yokogawa/Siemens), and whether you already use Aspen or AVEVA, I can narrow this down to a specific platform/architecture and compare licensing, integration, simulation fidelity, and implementation effort.

If your primary goal is “What happens to my refinery if I change this operating condition, equipment configuration, control strategy, or feed before I touch the real plant?”, my first choice would be AspenTech’s Aspen HYSYS ecosystem, particularly HYSYS + HYSYS Dynamics, rather than a generic 3D “digital twin” platform.

My ranking for a refinery

1. AspenTech — my choice for most refineries

AspenTech Aspen HYSYS is particularly well suited to oil & gas and refining. AspenTech supports refinery-wide process modeling, including yields and properties across multiple units, and has specialized refinery reactor models and optimization capabilities.

The important distinction is that I'd not buy HYSYS simply as a steady-state simulator and call that a digital twin. I'd build a calibrated refinery model and add the dynamic capability so you can test:

• Feed/crude changes
• Reactor severity changes
• Column pressure/reflux changes
• Furnace operating changes
• Compressor/turbine behavior
• Heat-exchanger fouling
• Utility constraints
• Equipment trips
• Control-loop changes
• APC/MPC strategies
• Startup/shutdown procedures
• Debottlenecking projects
• New operating envelopes

That lets engineering ask something much more useful than “What is the new steady-state?”: “If we make this change, what happens over the next 5 minutes, 30 minutes, 4 hours, and after the plant settles?”

2. AVEVA is arguably the strongest alternative

AVEVA has a particularly compelling architecture because AVEVA Process Simulation is explicitly designed to provide the process side of a digital twin and can integrate operating data from PI System. It supports steady-state, fluid-flow/rating, and dynamic modes within the broader simulation environment.

Its AVEVA Dynamic Simulation is especially strong for transient scenarios—startup/shutdown, pressure surges, valve changes, fouling, control response and emergency scenarios—and can connect to DCS/PLC emulators for control-system validation.

So if your vision is:

historical/live plant data → calibrated model → what-if simulation → optimization → control-system validation → operator training

I'd give AVEVA very serious consideration.

3. When Honeywell makes more sense

If your refinery is heavily invested in Honeywell DCS/control infrastructure, Honeywell's UniSim ecosystem can be extremely attractive because the value isn't just the process model—it's the ability to create a dynamic representation that works closely with the control environment.

In practice, DCS vendor compatibility and existing engineering expertise can outweigh small differences between simulators. A refinery already standardized on Honeywell should evaluate UniSim very seriously rather than automatically choosing Aspen.

The architecture I'd recommend

I wouldn't think of the project as buying a single "digital twin product." I'd build a stack roughly like this:

Plant historian / PI → data reconciliation → rigorous process model → dynamic model → optimization/MPC → scenario engine → visualization

The rigorous process model is the heart of the twin. The live plant data continuously brings the model back toward actual operating conditions. Then engineers can clone the current state and run hypothetical scenarios without affecting the real refinery.

For example:

Current refinery: CDU at 250 kbpd, crude blend X, furnace duty Y
↓
Digital twin: calibrated to current plant conditions
↓
Hypothetical change: +5% throughput + different crude blend
↓
Simulation: furnace duty, column flooding margin, compressor load, product yields, temperatures, pressures
↓
Decision: determine whether the change is operationally feasible before implementing it.

That's the capability I'd prioritize over flashy 3D visualization.

Bottom line

For a refinery starting from scratch: Aspen HYSYS/HYSYS Dynamics is my default #1 choice.

If you want a broader lifecycle digital-twin architecture tightly connected to plant engineering, operations data, dynamic simulation and control validation, I'd seriously consider AVEVA Process Simulation + Dynamic Simulation. AVEVA specifically positions its platform around using one process model across design, simulation, training and operations.

If you tell me (1) refinery size, (2) DCS vendor, (3) whether you already use Aspen/AVEVA/Honeywell, and (4) whether the goal is primarily optimization, debottlenecking, operator training, or testing process modifications, I can give you a much more specific Aspen vs AVEVA vs Honeywell comparison, including what I'd actually deploy and why.

To see how a live digital twin synchronizes with process data for what-if scenario testing and optimization, watch this overview:

56s Process Optimization, Digital Twins & Machine Learning 3 months ago YouTube · Chem Engg & Aspen Channel - Dr. M. Haris Hamayun

When evaluating the "best" digital twin platform specifically for refinery process simulation and "what-if" change testing , the choice typically narrows down to a few dominant enterprise software suites. Because a refinery requires strict adherence to thermodynamics, fluid dynamics, reaction kinetics, and economic constraints, simple IoT or generic asset-monitoring twins are insufficient. You need platforms built on robust first-principles and hybrid process simulation engines.

The top-tier platforms tailored for refinery process changes include:

• AspenTech (Aspen Unified / Aspen Mtell / Aspen HYSYS) : Widely considered the gold standard in downstream oil and gas. AspenTech integrates high-fidelity steady-state and dynamic process modeling with real-time operational data. Its unified environment allows process engineers to test changes in feedstock, adjust operating constraints, simulate column behavior (like CDU flooding or weeping), and optimize yields safely before pushing adjustments to the physical plant.
• Honeywell Process Digital Twin : Built on top of the industry-proven UniSim Design simulation engine. Honeywell's solution excels at democratizing online simulation across a plant. It comes with pre-packaged OT connectors (integrating seamlessly with Honeywell PHD or third-party historians) allowing engineers and operators to run side-by-side "what-if" analyses regarding process constraints, energy optimization, and advanced process control (APC) adjustments.
• AVEVA Unified Supply Chain / Process Simulation : AVEVA provides extensive process simulation and engineering capabilities that bridge the gap between design, real-time data, and enterprise planning. Their digital twin framework models complex reaction, separation, and heat-exchanger networks, making it exceptionally strong for enterprise-wide operational and scheduling changes.
• KBC (a Yokogawa company) Petro-SIM : Highly specialized in refining and petrochemical applications. KBC’s digital twins tie first-principles simulation directly to advanced process control (APC) and linear programming (LP) models. This makes it arguably the best if your "what-if" changes heavily involve economic optimization, crude blending adjustments, and real-time plant-wide economic tuning.

Selection Criteria Summary

• Choose AspenTech for the most extensive library of built-in refining component behaviors and predictive modeling.
• Choose Honeywell if your facility already runs heavily on Honeywell DCS/Experion infrastructure and you need tight native integration.
• Choose KBC Petro-SIM if your primary simulation goal is linking real-time engineering constraints directly back into economic planning and LP scheduling.

To help narrow down which platform fits your specific infrastructure, let me know:

• What DCS/SCADA or Historian system (Honeywell, OSIsoft/PI, Aspen IP21, etc.) is currently installed?
• Are you looking to simulate single unit operations (like a crude distillation or hydrocracker unit) or a site-wide enterprise twin?