Fouling is the most expensive problem in a refinery that nobody sees happening. The crude preheat train quietly loses heat-transfer performance as deposits build on the tubes, the crude reaches the furnace a little colder every week, and the furnace burns more fuel to make up the difference — so the cost surfaces as a rising fuel bill long before anyone connects it to a specific exchanger. In US refining alone, heat exchanger fouling in crude units was projected to cost on the order of a billion dollars a year. The trouble is that fouling gets managed on a calendar instead of on data, cleaning exchangers too early or too late. An AI fouling engine predicts each exchanger's fouling curve and schedules cleaning when it actually pays. You can book a demo to see it on your preheat train.
See Fouling as a Curve, Not a Fuel Bill — and Clean Each Exchanger When It Pays
Predict crude preheat train fouling, exchanger cleaning need, and thermal duty loss per exchanger — so you recover the energy fouling steals, defer cleaning spend until it's economically justified, and never let the furnace quietly burn the difference.
The Cost Shows Up at the Furnace, Not at the Exchanger That Caused It
The preheat train exists to recover heat from the crude unit's hot products and pumparounds and put it back into the cold crude before the furnace — every degree it adds is a degree the furnace doesn't have to. Fouling erodes that recovery slowly and silently: deposits accumulate on the tubes, the overall heat-transfer coefficient drifts down, and the crude arrives at the furnace colder. The furnace compensates automatically by firing harder, so the loss surfaces as fuel gas consumed, not as an alarm on an exchanger. By the time it's noticeable, weeks of energy are already gone.
As fouling cuts recovery, the crude reaches the furnace coil at a lower temperature. Every degree of lost preheat is a degree the fired heater must add with fuel — a direct, continuous cost that scales with throughput.
The furnace burns more fuel gas to hold the required crude temperature, driving up energy cost and carbon emissions together. This is the single largest cost of fouling, and it's the one that hides in the utilities budget.
Deposits narrow the flow path, so pumps work harder to push crude through the train. The added pumping cost is a real, often-ignored part of the fouling penalty — and accounting for it changes the optimal cleaning schedule.
When fouling is severe enough that the furnace hits its firing limit, the only way to hold coil outlet temperature is to cut crude rate — turning an energy problem into a lost-production problem worth thousands of barrels a day.
U-Value Is the Early Signal — If You're Actually Watching It
Fouling has a direct, measurable fingerprint before it ever reaches the fuel bill: the overall heat-transfer coefficient, the U-value, of each exchanger. A clean exchanger has a known U; as fouling builds, thermal resistance rises and U falls. Tracking the gap between clean and current U per exchanger — from the flows and temperatures the DCS already measures — is what turns fouling from an invisible drift into a visible, per-exchanger curve you can act on.
Using the inlet and outlet temperatures, flow rates, and duty the DCS already logs, the engine calculates each exchanger's current overall heat-transfer coefficient continuously — no new instrumentation, just the measurements already on the historian.
The difference between the clean-condition U and the current U is the fouling thermal resistance — the quantity that has been building silently. Trended over time, it becomes each exchanger's fouling curve, filtered from measurement noise using a least-squares approach so the signal is trustworthy.
The lost duty from each exchanger's fouling converts directly into the extra furnace fuel it's costing right now, so the fouling curve isn't an abstract resistance — it's a live dollar figure per exchanger that tells you which ones are actually hurting.
An AI model learns each exchanger's fouling behavior — which foul fast, which stay clean, how crude blend and rate change the rate — and projects the curve forward, so you know not just today's penalty but when each exchanger will cross the line where cleaning pays.
Turn Fouling Into a Live Dollar Figure Per Exchanger
iFactory computes each exchanger's U-value and fouling resistance from your existing DCS data and converts the duty loss into the fuel it's costing now — so you see exactly which exchangers are worth cleaning.
Clean Too Early and You Waste the Spend; Too Late and You Waste the Fuel
Cleaning an exchanger restores its heat transfer — but it isn't free, and it isn't costless to do. There's the direct cost of chemicals, labor, and equipment, and there's the hidden cost that taking an exchanger offline during operation removes surface area from the train, temporarily worsening recovery. That trade-off is exactly why cleaning on a fixed calendar leaves money on the table: the economically optimal moment to clean each exchanger is a genuine optimization, not a date.
Cleaning an exchanger that's only lightly fouled spends the chemicals and labor for little recovered duty, and takes surface offline for a marginal gain — capital burned ahead of when it was justified.
Cleaning each exchanger at the point where its accumulated fuel penalty has grown enough to justify the cleaning cost — and coordinated across the train — is where total avoided cost peaks. That's the schedule the engine solves for.
Letting an exchanger foul well past its optimum burns furnace fuel every day it waits, and risks pushing the train toward the throughput limit — the slow, silent version of the loss that's easiest to ignore.
The preheat train is a network — cleaning one exchanger changes the temperatures every exchanger downstream of it sees, so the exchangers are coupled and can't be optimized in isolation. And the objective has to include the hydraulic penalty, not just the thermal one: studies show that accounting for the added pumping cost of fouling, rather than heat loss alone, meaningfully changes the optimal schedule and raises the savings it captures. Solving that coupled, multi-exchanger, thermal-and-hydraulic problem is what an AI engine does that a maintenance calendar cannot.
Recover the Energy Fouling Steals, and Defer the Cleaning You Don't Need Yet
Everything the fouling engine does resolves into two financial outcomes that pull in the same direction: it recovers energy that fouling was quietly costing, and it defers cleaning capital until the moment it's justified. One grows the margin, the other protects the cash — and both come from seeing each exchanger's true economics instead of guessing.
By flagging the exchangers whose fouling is costing the most fuel now and getting them cleaned at the right time, the engine keeps coil inlet temperature high and furnace firing low — recovering energy that a calendar-based program leaves on the table for weeks.
Just as importantly, it stops you cleaning exchangers that don't need it yet. An exchanger fouling slowly can safely run longer than the calendar says, deferring the chemical, labor, and downtime cost until it's economically earned — real cash kept in the budget.
By keeping the train's recovery healthy, the engine keeps the furnace off its firing limit, so fouling never forces a crude-rate cut — protecting the marginal barrels that are worth far more than the fuel.
Every unit of furnace fuel avoided is carbon not emitted, so the same optimization that lowers energy cost lowers emissions intensity — a compliance and reporting win alongside the margin one.
The Same Preheat Train, Cleaned on a Date vs. on Its Economics
The difference between managing fouling on a fixed schedule and on a predictive engine shows up on every number that matters — fuel, cleaning spend, throughput, and how often you're cleaning the wrong exchanger.
| Decision | Calendar-Based Cleaning | AI Fouling Engine |
|---|---|---|
| Seeing fouling | Inferred late from the fuel bill | Live U-value curve per exchanger |
| Which exchanger to clean | Whichever is due by date | Whichever costs the most fuel now |
| When to clean it | Fixed interval, too early or too late | At its economic optimum |
| Pumping cost | Ignored | Included in the objective |
| Cleaning spend | Fixed, often premature | Deferred until justified |
| Throughput risk | Discovered at the firing limit | Protected by early warning |
Physics-Grounded, Data-Trained, and Honest About Noise
A fouling prediction only earns an operator's trust if it's grounded in the real heat-transfer physics and validated against how the train actually behaves. These are the foundations that make the engine's recommendations something a reliability team will act on.
The U-value and fouling-resistance calculations come from the thermal and hydraulic physics of shell-and-tube exchangers, so the numbers mean what they should — not a black-box correlation detached from how the equipment works.
On top of the physics, machine-learning models learn the fouling patterns specific to each exchanger and crude slate — which foul fast, how blend and rate shift the curve — the data-driven layer that makes the forecast accurate.
Process measurements are noisy, so the fouling signal is extracted with a least-squares approach rather than taken from raw readings, which is what keeps a spurious temperature blip from being mistaken for a fouling event.
Everything runs on the temperatures, flows, and pressures already on your DCS and historian, so the engine deploys without new field instrumentation — the signal to predict fouling is already being measured.
Per-Exchanger Fouling, a Live Cost, and a Schedule That Pays
iFactory's fouling engine reads your preheat train's existing data, computes each exchanger's fouling and its cost in real time, forecasts where the curves are heading, and solves the coupled cleaning schedule that maximizes avoided cost — turning fouling management from a calendar into an optimization.
What Refinery Reliability Teams Ask About Fouling Prediction
Stop Paying for Fouling at the Furnace
iFactory predicts crude preheat train fouling per exchanger, prices the duty loss as fuel, and solves the cleaning schedule that maximizes avoided cost — so you recover the energy fouling steals and defer the cleaning you don't need yet.
