AI-driven refinery process optimization
We build hybrid data-driven and physics-based models to monitor, predict and optimize process and equipment performance, turning plant data into actionable insights for a more efficient, profitable and sustainable future.
- Heat & material balance
- Simulation
- AI
- Digital twin
- Process optimization
- MonitorReal-time insights
- PredictPrevent issues
- OptimizeMaximize value
Refinery operations face growing data and efficiency gaps.
Incorrect / missing tags
Wrong or dead P&ID instrument tags distort plant data
Data inconsistency
DCS, LIMS and lab data rarely reconcile
Poor heat & material balance
Unclosed balances mask real performance
High utility consumption
Steam, power & fuel gas losses go undetected
Manual engineering calculations
Excel-only workflows are slow and error-prone
Time-consuming optimization
Simulation studies take days per case
Limited AI decision support
Little automated, data-driven recommendation
These gaps compound, costing energy, throughput and reliability every day.
An end-to-end, AI-enabled workflow.
A closed-loop pipeline, from raw plant data to actionable, self-tuning recommendations.
- 01Plant data
- 02Data validation
- 03Heat & material balance
- 04Tag normalization
- 05Simulation
- 06Optimization
- 07AI recommendation engine
- 08Continuous monitoring
Heat & material balance reconciliation.
Import process data
Pull live tags from DCS / IP21 / LIMS
Validate P&ID tags
Cross-check every instrument against the P&ID
Detect wrong instruments
Flag drifted, frozen or out-of-range signals
Normalize plant data
Reconcile flows, temperatures & pressures
Tune unknown parameters
Back-calculate missing or unmeasured values
Close mass & energy balance
Deliver a balanced, trustworthy dataset
Process flow, measurement points
- Feed preheat
- Fired heater
- Distillation column
- Product cooler
Highlighted nodes = reconciled flow, temperature & pressure tags feeding the balance model. The fired heater and the distillation column are covered in detail below.
Output: a fully closed mass & energy balance, the trusted foundation for every downstream calculation.
We built it properly for one unit before claiming it for a plant.
Fired heaters burn the largest share of a refinery’s fuel, so that is where we started. Everything on this page that is marked as built, we built here first: the physics, the soft sensors, the optimiser and the checks that catch it lying.
- The unit
- 53.4 MW absorbed, 195,000 kg/h of 34 API crude
- Solved to
- 90.7 % LHV at 15 % excess air on a clean coil
- Against
- All inside the API 560 bands
The stack temperature, the radiant and convection split, the tube metal temperature and the efficiency all fall out of the geometry and the energy balance. None of them is a correlation fitted to a known answer, which is why the model still behaves when it is pushed somewhere the data never went.
First-principles simulation of a 53.4 MW crude preheat heater driven through 120 days of hourly operation. Not a plant installation.
- Efficiency
- 87.1 %
- Stack O2
- 3.73 %
- Stack temp
- 233 C
- CO
- 20 ppm
- Feed
- 189,546 kg/h
- Outlet
- 415 C
- Fouling
- 0.75
- Detected fault
- fuel_upset
Trim excess air 19.8 % to 9.9 % (stack O2 3.73 % to 2.08 %)
$45.44/h~$381,710/yr at this state
- +1.26 ptsefficiency
- −1.52 %fuel
- −0.617 t/hCO2
Watch Move in steps and hold. Stop if CO passes 100 ppm before the O2 target, or if the stack falls below 98 C.
Why +−
Every percent of excess air above what the burners need is nitrogen heated from 20 C to the 233 C stack and vented. Cutting stack O2 by 1.65 points removes that parasitic mass flow, which is why the stack falls 16 C and dry flue-gas loss drops with it. The duty is unchanged, so the whole of the recovered loss shows up as less fuel through the burners. The move stops where it does because CO 167 ppm at the 200 ppm limit.
Reflux, reboiler duty and purity, balanced live.
The same physics-plus-AI method as the fired heater, applied to the column: find the operating window that holds both product specs at the lowest energy.
Key variables and relationships
- Feed rate
- Higher throughput drives production
- Reflux ratio
- Controls top product purity
- Reboiler duty
- Controls bottom product purity and energy use
- Product purity
- Top and bottom spec must stay in-window
- Energy consumption
- Minimized while meeting quality targets
Scenario engine: four simulation cases
- 01
Feed +10%
Predicts the reflux ratio and reboiler duty increase at higher throughput.
- 02
7-year composition sweep
Runs the historical feed composition range and tunes the model to real plant data.
- 03
Bottom spec fixed
Varies the top specification over its historical range and captures the deltas.
- 04
Top spec fixed
Varies the bottom specification over its historical range and captures the deltas.
AI rule engine: operating logic from live column signals
- Ifheavier goes upIncrease reflux ratio
- Iflighter goes downIncrease reboiler duty
- Ifpurity decreasesRecommend operating change
- Ifenergy exceeds targetSuggest optimal window
In development. The method is shown here; no column results are quoted until one has been validated against an operating unit.
Intelligent tag validation & auto-tuning.
- Identify incorrect transmitter tags
- Detect missing sensors
- Flag bad-quality measurements
- Reconcile data across sources
- Confidence score per tag
- Automatic parameter tuning
- FT-1042 Feed flowOK98%
- TT-2087 Column topDrift61%
- PT-3311 ReboilerOK95%
- FT-4456 RefluxBad quality34%
- TT-5502 BottomOK91%
Norms analysis: actual vs budget.
Every utility and raw material measured against its budget norm per unit of production, so an overrun is visible the week it starts.
- Utilities
+6.4%
Above budget
- Raw material
-1.1%
Within budget
- By-products
+0.8%
Within budget
- Energy / ton
+9.2%
Above budget
Example values from the platform design, not plant readings.
- Steam+6.7% (4.8 vs 4.5)
- Power+6.9% (62 vs 58)
- Fuel gas+6.9% (3.1 vs 2.9)
- Cooling water+7.7% (210 vs 195)
- Raw material-1.1% (98.9 vs 100)
Utility balance: OSBL to ISBL.
utilities
consumption
- Steam
- Cooling water
- Fuel gas
- Electricity
Hidden losses
Untracked leaks & venting
Startup consumption
Elevated draw during start-up
Miscellaneous loads
Unclassified auxiliary demand
Utility imbalance
OSBL supply vs ISBL demand gap
Power BI: utility & root-cause dashboard.
48.2
Steam (t/h)
1,860
Cooling water (m³/h)
3.1
Fuel gas (t/h)
12.4
Electricity (MW)
Example values from the platform design, not plant readings.
Steam +7% vs target (W5)
Traced to norms step 3: startup load
Cooling water spike
Traced to condenser fouling trend
Fuel gas above budget
Traced to heater excess air
From historical data to AI copilot.
Each stage compounds plant-specific intelligence: the model gets sharper with every operating cycle.
- Historical data
- Data cleaning
- Feature engineering
- AI model training
- Prediction
- Recommendation engine
- AI copilot
Future vision.
AI copilot
For every process engineer, every shift
Digital twin integration
Live virtual replica of plant units
Autonomous optimization
Self-adjusting operating windows
Predictive simulation
Forecast performance before it happens
Explainable AI
Transparent, auditable recommendations
Closed-loop optimization
Recommendations feed back automatically
Enterprise-wide deployment
Scaling across sites and units
From insight to autonomous action.
Business benefits.
The outcomes the full platform is designed to deliver once deployed on a plant.
Higher throughput
Operating windows set from a closed balance, so units can run closer to their real limits.
Lower energy use
Fuel and steam moves priced in money before anyone changes a setpoint.
Product quality consistency
Soft sensors fill the gaps between lab samples, so drift shows up sooner.
Reduced emissions
CO2 tracked per unit of feed, next to the fuel it comes from.
Faster troubleshooting
Drifted and frozen tags flagged before they mislead a diagnosis.
Improved profitability
Each recommendation carries its value in money, so the payback is visible.
Better data quality
Tags checked against the P&ID and reconciled before any model reads them.
Built for today’s challenges. Ready for tomorrow.
Our AI-powered digital twin solutions help you achieve measurable results, now and in the future.
Energy and utilities
Improve efficiency, reduce energy use
Oil, gas and refining
Maximize uptime, optimize operations
Chemicals and petrochemicals
Stabilize processes, ensure product quality
Power and renewables
Enable cleaner, smarter operations
Fertilizers and specialty chemicals
Increase yield, reduce costs
One plant, six layers. Three of them exist today.
The end state is every unit on a site modelled, reconciled and optimised continuously. We are building it in the order that makes each stage useful on its own, and we mark honestly which stages are finished.
- 01In progress
Heat and material balance
Reconcile flows, temperatures and pressures until mass and energy close. Nothing downstream is trustworthy until this does, and in most plants it has never been done on live data.
- 02Planned
Tag validation
Find the instruments that are drifted, frozen, mis-mapped or simply wrong, and score the confidence of every tag before a model is allowed to read it.
- 03Built
First-principles simulation
Build the equipment properly: combustion, heat transfer, hydraulics and an efficiency ledger that a process engineer can audit line by line.
- 04Built
Constrained optimisation
Find the best safe operating point against real limits, price the move in fuel, CO2 and money, and name the constraint that stops it going further.
- 05Built
Machine learning layer
Soft sensors for the state nobody measures, a surrogate fast enough to search a year of operation, and anomaly detection that does not depend on fault labels.
- 06Planned
Operator copilot
Every unit on the site, answering questions in plain language and carrying the reasoning of engineers who have retired.
3 of 6 phases built and tested, for one unit type. Everything marked in progress or planned is a statement of direction, not a capability you can buy today.
Reliable data. Smarter simulation. Autonomous optimization.
- A validated heat & material balance is the trust layer for every decision
- Simulation + AI rules turn plant data into repeatable, tunable operating logic
- A phased MVP delivers value at every stage, toward a full AI copilot
Questions & discussion
Send us 90 days of tags from one unit. We will close the balance, validate the tags and show you what it is leaving on the table.