PCI–RAFT Digital Twin
Interactive plant-facing tool for blast furnace PCI evaluation. The twin compares thermochemical and empirical RAFT predictions, tracks coke and total reductant rate, and shows CFD-informed raceway profiles for engineering interpretation.
RAFT Model Selection
AISI / Sheth is the default broad-range empirical comparator for conventional PCI operation.
Blast & Injectant Parameters
Coal, Coke & Burden
RAFT Gauge —
Target operating window: 1950–2250 °C. Below this range, combustion stability and burden melting become weaker; above it, refractory load and silicon pickup become harder to control.
RAFT Comparison
The upper chart compares thermochemical and empirical RAFT. The lower chart shows the term-wise breakdown of the selected empirical model.
Coke Rate Contribution Breakdown
This breakdown shows the directional effect of PCI substitution, blast enthalpy, oxygen enrichment, burden Fe, flux, top pressure, silicon target, and coke quality on the calculated coke rate.
How RAFT Is Calculated
The raceway adiabatic flame temperature is the theoretical lower-zone gas temperature obtained when the net chemical and sensible heat released around the tuyere is converted into the sensible heat of the hot product gas.
Empirical RAFT
The empirical side of the twin provides fast operational estimates. Three options are included: the AISI / Sheth correlation for general PCI use, a simpler legacy operator formula, and a hydrogen-rich-gas comparator intended only for special co-injection cases.
The empirical models are useful because they react quickly and transparently to operating changes, making them suitable for plant control-room interpretation.
Thermochemical RAFT
The thermochemical side is implemented as a lower-zone enthalpy framework anchored to the empirical base but corrected with physically directional thermal terms representing hot-blast sensible heat, oxygen enrichment, blast moisture, PCI cooling, fuel moisture, ash, coke quality, burden quality, flux load, and silicon target.
This gives a stable, plant-usable thermochemical estimate while still preserving the correct directional influence of the main tuyere-zone and burden variables.
Engineering interpretation
The empirical and thermochemical values should be used together. When both move in the same direction, the thermal interpretation is robust; when they begin to diverge materially, the deviation should trigger plant review of blast chemistry, fuel assumptions, burden quality, or lower-zone calibration constants.
Coke Rate & Reductant Balance
Coke rate is calculated from PCI substitution, burden quality, blast thermal state, and slag-driving variables. The twin uses these terms to estimate both coke rate and total reductant rate.
Coke-rate model
The equation is intended as a practical operating model. It captures the expected thermal and burden-quality directions without claiming full furnace closure.
Replacement ratio
Higher carbon and hydrogen raise replacement ratio, while higher moisture penalizes it. Ash is treated as a weak positive term here only as a fitting balance term and should be recalibrated with plant data if needed.
Coke–PCI Trade-off
Parameter Sensitivity Studies
Each sweep varies one parameter while all others remain fixed at the selected operating point.
Blast Temperature
Blast Moisture
Oxygen Enrichment
PCI Rate
Natural Gas / H2-rich Fuel
Coal Ash
Coke Moisture
Coke CSR
Required Blast Temperature to Hold RAFT Constant
Coke Rate vs PCI at Different O₂ Levels
Coal & Raw Materials
Coal rank, ash, volatile matter, moisture, fineness, coke quality, burden Fe, flux, and target thermal state all influence the sustainable PCI rate.
Coal-rank impact
| Coal type | VM range | Typical RAFT effect | Combustion behaviour | Typical RR |
|---|---|---|---|---|
| Low volatile | 10–18% | Lower cooling penalty | Slower ignition, stronger char control | 0.75–0.85 |
| Medium volatile | 18–28% | Balanced | Moderate ignition and burnout | 0.85–0.92 |
| High volatile | 28–40% | Larger cooling penalty | Faster gas-phase ignition | 0.90–1.00 |
Operational interactions
- Higher coal ash increases the heat burden and usually lowers RAFT and raises slag-related load.
- Higher burden Fe lowers gangue burden and tends to improve both coke rate and furnace thermal efficiency.
- More raw flux raises the in-furnace heat requirement and generally pushes coke rate upward.
- Poor burden permeability or weak coke quality limits the usable PCI rate long before the nominal thermal model limit is reached.
- Hot metal silicon acts as a practical thermal-state proxy and is therefore included as a load term in the coke-rate model.
CFD-Informed Raceway Profile
This section provides a lower-zone profile layer shaped from accepted raceway behaviour. It is intended for interpretation and digital-twin coupling, not as a substitute for a full CFD solve.
Included physics
- Gas-phase reacting flow with hot blast, oxygen enrichment, and moisture effects
- Coal-particle heating, devolatilization, char oxidation, and burnout progression
- Porous coke-bed interaction at the edge of the raceway
- High-temperature gas mixing and species conversion in the tuyere zone
- Radiative and convective lower-zone heat transfer interpretation
What this profile means
The plots in this section are CFD-informed profiles tied to the thermal state selected in the twin. They are not a live Navier–Stokes solution but a structured engineering profile layer for plant use and later calibration.
Axial temperature profile
Species profile
Burnout & velocity
Raceway depth sensitivity
How to turn this into a real CFD-coupled twin
- Build a validated tuyere-raceway CFD sector for the furnace.
- Run a designed set of cases across PCI, O₂, blast temperature, humidity, fuel type, and coke-bed permeability.
- Extract bulk RAFT, peak temperature, species, burnout, raceway depth, and char carryover.
- Train a surrogate model on the CFD results.
- Replace or calibrate the present profile functions with the trained surrogate.
Recommended deployment structure
| Layer | Typical implementation |
|---|---|
| Plant data | Historian / PLC / Level-2 feed |
| Fast process model | This browser twin or a Python service |
| High-fidelity layer | OpenFOAM, Fluent, or STAR-CCM+ |
| Surrogate | Python regression or ML model |
| Operator UI | Dashboard, historian overlay, or Level-2 advisory tool |
Methodology & Boundaries
What was corrected
This revised twin removes visible source tags, corrects inconsistent plot behaviour, rebuilds the RAFT comparison logic so that it follows the selected empirical model, and fixes the blast-temperature compensation calculation so the sensitivity plot behaves correctly.
The CFD section remains intentionally labelled as CFD-informed rather than a live CFD solver. Its role is to provide a realistic profile layer and a path toward future calibration with actual raceway simulation results.
For plant trials, the next step should be calibration against site-specific coke rate, hot metal silicon, tuyere-gas chemistry, and any available lower-zone or tuyere camera interpretation.