AI Desalter Optimization for Crude Distillation: Improve Salt Removal & CDU Performance

By Johnson on September 2, 2026

ai-desalter-performance-optimization-crude-distillation

A crude unit runs for weeks with desalter salt content sitting comfortably under target, and then a single crude slate change quietly pushes it past the line. Nobody notices until preheat exchangers start fouling faster than the cleaning schedule accounts for, or a corrosion coupon in the overhead comes back showing chloride attack nobody budgeted for. The desalter did not fail; the mix valve pressure, chemical dose, and interface level that used to be right for one crude blend were simply never re-tuned for the next one, and by the time a lab result flags it, the damage is already downstream. You can see how continuous AI tuning closes that gap by choosing to book a demo with our team.

CRUDE DISTILLATION · DESALTER OPTIMIZATION · SALT CARRY-OVER CONTROL

Your Desalter Is Tuned for the Crude You Ran Last Month, Not the Crude You're Running Today

Mix valve pressure, chemical injection rate, temperature, and mud wash all interact, and the right setpoints for one crude blend can be exactly wrong for the next. iFactory continuously re-tunes desalter operating parameters against live feed characteristics so salt carry-over stays inside target instead of drifting until a corrosion or fouling event forces a correction.

SALT CONTENT
1.4 PTB
Target: below 1.5 PTB
MIX VALVE ΔP
8.2 PSI
Within optimal shear range
INTERFACE LEVEL
Stable
No rag layer growth flagged
WHY SALT CARRY-OVER IS SO HARD TO CATCH IN TIME

The Desalter Sits Between Four Variables That All Move at Once

A desalter does one job: strip enough water-soluble salt and suspended solids out of crude before it reaches the atmospheric tower, so downstream chloride hydrolysis does not generate hydrochloric acid in the overhead. The mechanism for doing that job well depends on wash water mixing thoroughly with the crude, which depends on mix valve pressure drop, which depends on crude viscosity, which changes with every blend and every ambient temperature swing.

Operators typically set mix valve pressure once, based on the crude the unit was designed to run, and leave it there through blend changes that were never modeled into that original setpoint. Too little shear and the wash water never contacts enough of the salt; too much shear and the emulsion becomes so stable it will not separate in the settling section, which shows up as a rising rag layer and eventually as oil-in-water carry-under or water-in-oil carry-over, either direction of failure.

Chemical demulsifier dose compounds the same problem. A dose rate tuned for a heavier, more asphaltenic crude will often overtreat a lighter blend, which can itself destabilize the interface rather than help it, while a dose tuned for a light sweet crude will consistently undertreat a heavier opportunity crude brought in to capture a margin advantage. Because demulsifier chemistry is crude-specific and blend ratios shift week to week on many units, a fixed dose curve is almost never the right dose curve for more than a portion of the operating window.

1.5 PTB
Common salt content target at the desalter outlet, pounds of salt per thousand barrels of crude
90-95%
Typical salt removal efficiency achievable from a properly tuned single-stage desalter
3 PPM
Approximate neutralizer demand increase in the overhead per 1 PTB of salt carry-over above target
THE FOUR LEVERS AI ACTUALLY TUNES

Temperature, Mix Valve Pressure, Chemical Dose, and Mud Wash Have to Move Together

Treating these four variables as independent settings is the most common reason a desalter optimization effort stalls. They interact, and the correct value for any one of them depends on where the other three currently sit relative to the crude blend actually flowing through the vessel right now.

01
Desalter Temperature
Higher temperature lowers crude viscosity and improves droplet settling velocity, but pushing temperature too high near the crude's flash point risks vaporization in the vessel. AI holds temperature at the highest stable point the current blend supports.
02
Mix Valve Pressure Drop
Controls shear and therefore droplet size of the wash water dispersion. AI recalculates the optimal ΔP band continuously against live viscosity and water cut, rather than holding a single fixed setpoint across every blend.
03
Chemical Injection Rate
Demulsifier dose is matched to the asphaltene and resin content implied by the current blend ratio, cutting overtreatment on light crudes and closing gaps on heavier opportunity crudes before carry-over shows up in a lab sample.
04
Mud Wash Frequency
Solids accumulation at the bottom of the vessel reduces effective residence time and can bridge the electrodes. AI schedules mud wash cycles based on accumulation trend rather than a fixed calendar interval.

See What Your Current Desalter Setpoints Are Actually Costing You

iFactory maps live mix valve pressure, chemical dose, and salt content trends against your blend schedule so drift gets corrected before it reaches the tower.

FIXED SETPOINTS VERSUS CONTINUOUS RETUNING

What Changes When the Desalter Adjusts to the Blend Instead of the Other Way Around

Most units already have the instrumentation needed to run this way, salt-in-crude analyzers, mix valve ΔP transmitters, interface level detection, and dosing pumps, but the control logic connecting them was set once and rarely revisited. The gap is not equipment, it is the absence of a system that recalculates the right combination of setpoints every time the feed changes.

Operating Approach Fixed Setpoint Operation AI-Tuned Operation
Response to Blend Change Same mix valve ΔP and dose regardless of crude switch Setpoints recalculated against live viscosity and water cut
Chemical Cost Dose set for worst-case crude, overtreating lighter blends Dose matched to current asphaltene and resin demand
Mud Wash Timing Calendar-based interval regardless of solids accumulation Triggered by accumulation trend, not a fixed schedule
Detection of Salt Excursion Discovered on next scheduled lab sample Flagged from live analyzer trend before target is exceeded
A FOUR-PHASE ROLLOUT

Getting From Fixed Setpoints to Continuous Retuning

1
Baseline the Current Operating Window
Pull historical salt content, mix valve ΔP, dose rate, and crude assay data to establish where setpoints have actually been drifting relative to blend changes over the past operating cycles.
2
Connect Live Analyzer and Dosing Data
Integrate salt-in-crude analyzers, mix valve transmitters, interface level instruments, and chemical dosing pumps into one continuously updated operating picture.
3
Run Recommendations in Advisory Mode
AI-calculated setpoint adjustments are surfaced to operators as recommendations first, building confidence before any closed-loop control is considered.
4
Tighten the Salt Content Control Band
As the recommendation accuracy is validated against lab results, the operating band around target salt content is progressively tightened.
COMMON MISTAKES

Where Desalter Optimization Efforts Usually Go Wrong

Tuning Chemical Dose in Isolation
Increasing demulsifier dose without also checking mix valve shear and temperature often masks a mechanical setpoint problem instead of fixing it, and adds ongoing chemical cost that never needed to be there.
Ignoring Blend Ratio Changes From Procurement
Crude purchasing decisions made for margin reasons frequently reach the desalter without any corresponding review of whether current setpoints still fit the new blend composition.
Fixed Mud Wash Intervals on Variable Solids Crudes
A calendar-based mud wash schedule works fine until a high-solids opportunity crude is run, at which point accumulation between washes can exceed what the fixed interval was ever designed to handle.
Waiting for the Lab Result to React
By the time a scheduled salt content sample comes back showing an excursion, the crude that caused it may have already moved through the tower, leaving no window to correct it before it becomes a corrosion event.
MATCHING CHEMISTRY TO THE CRUDE ACTUALLY IN THE VESSEL

Demulsifier Selection Is a Second Layer of the Same Problem

Dose rate is only half of the chemical picture. The demulsifier chemistry itself, its solvent package, its HLB balance, and its treating speed, is formulated around a specific range of asphaltene, resin, and wax content, and a product selected for a heavier opportunity crude can underperform badly on a lighter blend even at a correctly scaled dose. Refineries that run a wide blend rotation often standardize on a single demulsifier for operational simplicity, which trades chemical performance across part of the rotation for procurement convenience.

AI-assisted blend tracking makes it possible to flag when the current feed has drifted far enough from the chemistry's design window that a product switch, rather than a dose adjustment, is the more effective correction. This does not mean carrying a large chemical inventory; it means having visibility into when the existing product is being asked to do a job it was not formulated for, so the decision to switch or blend chemistries is made deliberately rather than discovered through a rag layer problem weeks later.

WHO OWNS EACH PART OF THE LOOP

Desalter Optimization Crosses Three Roles That Rarely Share a Dashboard

Setpoint drift is as much an organizational gap as a technical one. Crude purchasing decisions, board operator setpoint adjustments, and process engineering review of salt content trends are typically owned by three different people who see three different slices of the same problem, and none of them individually has the full picture needed to catch a drift before it becomes an excursion.

Board Operator
Executes setpoint changes in real time and is the first to see mix valve ΔP and interface level trends, but rarely has direct visibility into the upcoming blend schedule from procurement.
Process Engineer
Reviews salt content and chemical consumption trends against target, and is best positioned to authorize a demulsifier chemistry change, but typically works from periodic reports rather than a live feed.
Crude Purchasing
Sets the blend schedule that drives everything downstream, but is rarely looped into whether the desalter is currently operating with comfortable margin or is already close to its limit.
FREQUENTLY ASKED QUESTIONS

Questions Process Engineers Ask About Desalter AI Optimization

Does this require new desalter hardware or instrumentation?
In most cases, no. Units already running salt-in-crude analyzers, mix valve ΔP transmitters, and dosing pumps have the instrumentation needed, the gap is usually in the control logic connecting that data to setpoint decisions rather than in the sensors themselves. A quick instrumentation review during onboarding confirms what is already in place. Book a demo to walk through what your current instrumentation already supports.
How quickly can setpoint recommendations respond to a blend change?
Recommendations update as soon as live viscosity, water cut, and analyzer trend data reflect the new blend, which is typically well ahead of when a scheduled lab sample would surface the same shift. Advisory mode is used first so operators can validate the recommendation against their own judgment before any closed-loop adjustment is considered. This closes most of the lag that fixed setpoints carry through a blend transition.
Can this reduce demulsifier chemical spend?
Yes, in most fixed-dose operations the dose curve is set to handle the heaviest crude in the rotation, which means lighter blends are routinely overtreated. Matching dose to the asphaltene and resin demand of the current blend typically reduces average chemical consumption while holding or improving salt removal performance. Contact our support team to review your current dosing curve against blend history.
What happens if mix valve pressure is set too high?
Excess shear creates a finer, more stable emulsion that resists separation in the settling section, which shows up as rising rag layer thickness and eventually as carry-under or carry-over in either direction. This is one of the more common mistakes in manual tuning, since more shear intuitively seems like it should mean better mixing, when in practice there is a defined optimal band specific to the current crude's viscosity and interfacial tension.
How does this affect downstream corrosion risk in the overhead?
Salt carry-over above target increases chloride availability for hydrolysis reactions in the overhead system, raising hydrochloric acid formation and neutralizer demand. Holding salt content consistently within target reduces that chloride loading at the source rather than relying entirely on overhead neutralization and corrosion monitoring to manage the consequence. Book a demo to see how salt content trends are tracked against overhead corrosion indicators.

Stop Re-Tuning the Desalter After the Blend Has Already Changed

iFactory keeps mix valve pressure, chemical dose, temperature, and mud wash timing matched to the crude actually running through the vessel, not the crude the unit was last tuned for.


Share This Story, Choose Your Platform!