Where the cost of getting it wrong
is measured in barrels per day

Refineries and hydrocarbon processing facilities are among the most data-rich environments in industry — and among the least digitally integrated. The gap between what existing instrumentation can tell you and what operations teams actually see is where margin leaks quietly, at scale, every day. Karnex focuses on closing that gap through calibrated digital intelligence, without overpromising what AI can do in a complex, high-hazard environment.

1.5–3%

Of gross throughput value recoverable through integrated digital twin deployment

up to 30%

Reduction in unplanned downtime from predictive maintenance at comparable facilities

Year 1

Typical payback horizon for data integration and production accounting phases

// The Problem

Most refinery losses are invisible on current dashboards

Downstream hydrocarbon processing operates with a persistent set of structural inefficiencies that conventional reporting does not surface. These are not operational failures — they are the predictable consequence of data architectures that were never designed to support integrated optimisation.

The Delta Vector Problem

The gap between metered, planned, and actual production — what the industry calls the "delta vector" — represents real value that is systematically unaccounted for. At large complex refineries, production accounting reconciliation alone typically recovers 0.1–0.3% of throughput value that was previously invisible in the data.

Conservative Manual Operation

Without Advanced Process Control, experienced operators run processes conservatively — well inside operating envelopes — to avoid trips and upsets. This is rational behaviour, not poor practice. But it leaves 0.5–2 percentage points of crude throughput assigned to low-value products that APC and RTO deployment would capture as gasoline or distillate.

Unplanned Corrosion Events

In facilities processing sour, high-sulfur crude, corrosion-driven failure is not a low-probability tail risk — it is a documented, recurring cause of unplanned downtime. A single multi-day shutdown event on a large processing unit represents throughput exposure that can exceed the total cost of a predictive maintenance programme deployed across an entire facility.

Brownfield-Greenfield Integration Gaps

Major refinery upgrade programmes routinely commission new high-conversion units alongside legacy distillation trains with decades of continuous service. The resulting hybrid — two very different process realities, different instrumentation vintages, different failure modes — is precisely the configuration where integrated digital intelligence delivers the most value and where ad-hoc, unit-by-unit approaches fall shortest.

// Structural Reality

A complex refinery processing heavy sour crude is plausibly leaking the equivalent of 1.5–3% of throughput value annually through unplanned downtime, suboptimal yield allocation, and energy inefficiency that a calibrated digital twin directly targets. At scale, this is not a marginal efficiency question — it is a material operating cost.

// Application Areas

Six lenses. One integrated view of the plant.

Refinery digital twin programmes fail when they are built as technology deployments rather than as operational capability. Each application area below is sequenced by data readiness and value delivery — not by what is technically most interesting to build first.

// 01 — FOUNDATION

KPI Visualisation & Benchmarking

Real-time performance gap surfaced between actual operations and design-basis benchmarks — at unit level and refinery level simultaneously. This is the enabling layer for every application that follows. It converts raw historian data into operational intelligence that process engineers and plant managers can act on without waiting for a monthly report.

Benchmark: 0.2–0.5% throughput-equivalent from faster fault detection alone

// 02 — RECONCILIATION

Production Accounting & Data Reconciliation

Systematic closure of the gap between metered, planned, and actual production. Connected to existing ERP and DCS historian systems, production accounting reconciliation gives finance and operations a single version of the truth — and typically surfaces inventory losses and mis-allocations that were previously invisible to both. This phase delivers return in Year 1 with contained integration scope.

Benchmark: 0.1–0.3% of throughput typically recovered from reconciliation alone

// 03 — OPTIMISATION

LP Model Updating & Supply Chain Economics

Refinery linear programming models used for crude selection and product-slate planning degrade in accuracy as actual unit yields diverge from design-basis assumptions — which happens immediately post-commissioning and accelerates as units age. Keeping the LP model calibrated against a live digital twin changes crude selection decisions and, at large-scale facilities, delivers measurable margin uplift per barrel processed.

Benchmark: USD 0.3–0.8/bbl margin uplift at comparable complex refineries

// 04 — CONTROL

Advanced Process Control (APC) & Real-Time Optimisation (RTO)

APC stabilises individual unit operations closer to their constraint limits, reducing process variability and shifting yield toward higher-value light products. RTO coordinates across units — crude preheat, primary distillation, conversion units — to optimise the refinery as an integrated system rather than a collection of independently managed processes. This is where the largest absolute value accrues, and where vendor and instrumentation choices have the most consequence.

Benchmark: 0.5–2 percentage point yield shift toward light products; 1–3% energy savings

// 05 — INTEGRITY

Predictive Maintenance & Corrosion Monitoring

The digital twin's process model, combined with corrosion-rate modelling tracking wall-thickness degradation against process conditions — temperature, velocity, H₂S and naphthenic acid content of the crude blend — converts corrosion-driven failure from a reactive problem into a planned, scheduled intervention. For facilities processing sour crude through aged equipment, this is frequently the highest-expected-value application in the portfolio, and the one most consistently underweighted in initial programme scoping.

Benchmark: up to 30% reduction in unplanned downtime; up to 25% reduction in maintenance OPEX

// 06 — SAFETY

Operator Training Simulator (OTS)

Dynamic simulation models built for process optimisation double as high-fidelity training environments. For high-hazard units — hydrocrackers, hydrogen production, large-scale crude distillation with complex heat integration — an OTS built on the same calibrated model used for RTO allows operators to rehearse start-up, shutdown, and emergency scenarios without risk to the live plant. This is especially material in the months following commissioning of new high-conversion units, when crew familiarity is lowest and consequence of error is highest.

Application: risk mitigation on newly commissioned high-hazard units; standard of care at comparable facilities

// Cost of Inaction

The decision is not whether to spend — it is whether to recover

The correct framing of the investment decision is not "commit capital to deploy a digital twin." It is "continue losing recoverable value, or recover it." Four categories of cost accrue for every year a programme does not begin.

Compounding Value Leakage

At the conservative low end of benchmark ranges, a complex sour-crude refinery leaks approximately 1.5–3% of throughput value annually through inefficiencies a digital twin directly targets. This is not a one-time loss — it compounds. After three years of inaction, cumulative foregone value routinely exceeds the total cost of the programme. The opportunity cost of delay is larger than the programme investment at any point after Year 1.

Accelerating Corrosion Risk Without Monitoring

Without corrosion-rate monitoring, the timing of the next failure event on aged high-sulfur-crude equipment is reactive, not predictive. The failure probability does not stay constant — it increases with continued operation. The financial exposure of a single unplanned multi-day shutdown on a large processing unit is not marginal. The predictive maintenance component of a digital twin programme typically costs a fraction of a single avoided event.

New Capital Running Below Its Optimum

Major refinery upgrade programmes commission new high-conversion units without an APC layer, without a calibrated LP model, and without an RTO coordinator. In the absence of these tools, experienced operators run new equipment conservatively — and continue to do so for years, because no optimisation programme was deployed. Each month of operation below the APC-enabled optimum represents foregone margin on the new capital investment.

Divergence from Regional and Global Peers

Tier-1 refining operators globally have mandated digital twin as standard for capital projects. For every year a facility does not deploy a digital twin, it falls further behind the operational practice of its peers — and the gap is compound in nature, because those peers are continuously improving their programme maturity. This is not primarily a competitive argument; it is a capability-building argument. The internal expertise required to run a world-class digital refinery takes years to develop.

// Implementation Sequencing

Where you start inside a facility determines what you can build

Programme sequencing across units within a refinery follows three criteria applied in order: data readiness (what instrumentation is actually usable), throughput leverage (how much of the facility's total value flows through this unit), and risk profile (what happens if the model is wrong for this unit). The right starting point is not the most technically interesting unit — it is the one that enables the fastest calibrated value delivery while establishing the data foundation for the rest of the programme.

Primary Distillation — New or Recently Commissioned Train

Modern instrumentation, documented calibration histories, and known DCS protocols make recently commissioned crude distillation units the right anchor for Phase 0 data integration and Phase 2 APC deployment. At most complex refineries, the primary CDU represents the largest single throughput leverage point in the plant. The calibrated model of the new unit also becomes the reference benchmark against which the degraded performance of older trains is measured.

Phase 0 / 1 / 2 APC Primary Target Highest Throughput Leverage

High-Conversion Units — Hydrocracker / RFCC

High-conversion units operate at elevated temperature and pressure, making them the highest-hazard processes in the facility and the most consequential for operator training. The Operator Training Simulator priority for these units is not a secondary consideration — it is the primary digital twin application. APC on high-conversion units delivers substantial yield and energy value, but is sequenced after OTS deployment once operator familiarity is established.

OTS Priority Phase 2 APC Safety-First Sequencing

Legacy Distillation Trains — Corrosion Monitoring Focus

Older distillation units processing sour crude carry the highest corrosion risk and the highest single-event downtime exposure. They are not the right starting point for digital twin deployment, however, because their data quality is the most degraded. The correct sequencing is to use Phase 0–1 to audit what data is actually usable on legacy trains, use the calibrated new-train model as a reference point, and begin corrosion-rate modelling in Phase 3 once a baseline of cleaned, validated data is established.

Phase 3 Target Predictive Maintenance Data Remediation First

Crude Preheat Train & Heat Exchanger Network

The crude preheat heat exchanger network is universally the largest single energy-efficiency lever in a sour-crude refinery. Progressive fouling of heat exchangers — from asphaltene and inorganic deposit accumulation — increases furnace firing rate and fuel consumption in direct proportion to heat-transfer degradation. A digital twin tracking heat-transfer coefficients in real time against clean-conditions design baseline identifies which exchangers are fouling fastest and should be cleaned, rather than following a fixed schedule regardless of actual condition. Temperature and flow measurements across the preheat train are typically already available in the historian — making this a relatively contained, high-return parallel workstream for Phase 1.

Phase 1 Parallel Energy Quick-Win 1–3% Fuel Savings

// Why Karnex

Analysis that comes from having built it

The intelligence Karnex publishes on oil, gas, and refining comes from the same depth of knowledge we bring to large-scale implementation. That distinction matters in a sector where the gap between theoretical optimisation and operational reality is wide and consequential.

// Domain Depth

Process Engineering, Not Platform Advocacy

Our analysis of digital twin and APC/RTO deployment is grounded in process engineering — steady-state and dynamic simulation, LP modelling, corrosion chemistry — not in any single vendor's platform architecture. We cover the real constraints: data quality, instrumentation vintage, calibration drift, organisational change, and the difference between what a platform claims and what it delivers under brownfield conditions.

// Operational Realism

Brownfield Complexity Is Our Default Assumption

Most published digital twin content assumes a greenfield instrumentation baseline and a cooperative DCS vendor. Refinery reality involves 70-year-old distillation trains with undocumented sensor histories sitting alongside newly commissioned high-conversion units on the same site. We write and advise for that reality, not for the reference architecture on a vendor's slide deck.

// Geography

MENA, European, and Emerging Market Context

Refining capacity expansion and upgrade programmes are concentrated in MENA, Sub-Saharan Africa, and South Asia — markets with different technology-sourcing constraints, domestic capability profiles, and regulatory environments than the Western European refinery digital twin literature typically assumes. Karnex's intelligence is calibrated for this geography, not for a North Sea or US Gulf Coast reference frame.