Sand does not care about your production forecast. A well that has produced clean for three years can start lifting fines the week after a workover, and by the time a choke shows a visible drop in performance, the erosion behind that drop has usually been building quietly for months inside a bend, a tee, or a valve trim nobody was watching closely. Sand erosion is one of production engineering's oldest problems, but predicting exactly where erosion damage will concentrate and how fast a specific fitting is wearing has traditionally required either a slow, expensive CFD study or a maintenance team's gut feel built from years on the same field. To see how erosion prediction would score against your current sand exclusion and desanding setup, book a session with the iFactory production integrity team.
Flow Assurance Intelligence · Erosion & Sand Management
AI for Erosion Prediction and Sand Management in Production Pipelines
Model sand production rate, erosion velocity, and choke valve wear together to predict damage before it happens — then optimise sand exclusion screen design and desanding equipment operation around the forecast.
How Erosion Actually Builds
Four Stages Between Clean Production and a Failed Fitting
Erosion damage rarely announces itself with a single event — it accumulates through a sequence that is well understood in principle but hard to track continuously without a model tying the stages together.
01
Sand production begins
Fines are lifted from the formation as flow rate, drawdown, or completion condition changes, often without a visible surface indicator until an acoustic sand detector or periodic sand cut test flags it.
02
Particle velocity concentrates at geometry changes
Bends, tees, reducers, and choke trim accelerate particle velocity locally, creating erosion hotspots that wear far faster than the straight-run pipe sections around them.
03
Wall thickness or choke trim thins progressively
Metal loss accumulates at the hotspot, gradually reducing the safety margin against the line's design pressure rating — a trend visible in erosion probe or choke position data long before failure.
04
Failure or unplanned shutdown
Without intervention, thinning eventually reaches a point requiring emergency choke replacement, unplanned line isolation, or in severe cases a loss-of-containment event.
Building the Prediction
What an AI Erosion Model Correlates to Score Risk Per Fitting
Sand Production Rate
Acoustic sand detector output and periodic sand cut test results, trended over time rather than read as isolated snapshots.
Erosion Velocity
Calculated against industry-standard erosional velocity guidance and adjusted for actual fluid density and particle characteristics rather than a generic assumption.
Geometry Risk Factor
Each bend, tee, reducer, and choke is weighted by its known erosion multiplier relative to straight pipe, based on established erosion modelling research.
Erosion Probe and Wall Thickness History
Ultrasonic thickness readings and intrusive probe trends provide the ground truth the model calibrates its predictions against for each specific fitting.
Choke Position and Wear Trend
Rate of change in choke position relative to a smoothed production rate curve is a strong early indicator of choke trim erosion, consistent with findings from real-time choke health monitoring research.
The Practical Difference
Scheduled Inspection vs Continuous Erosion Risk Scoring
| Programme Element | Scheduled Inspection | Continuous AI Scoring |
|---|---|---|
| Erosion detection timing | Found at the next periodic inspection interval | Trend flagged as soon as risk score crosses threshold |
| Highest-risk fittings identified | Inspected on the same schedule as low-risk fittings | Prioritised by calculated erosion risk score |
| Sand exclusion screen sizing | Set at design and rarely revisited | Reviewed against actual observed particle size and rate trends |
| Desanding equipment operation | Run on a fixed cycle regardless of sand load | Adjusted to match forecast sand production rate |
See Your Highest-Risk Fittings Ranked
iFactory Correlates Your Sand Detectors, Erosion Probes, and Choke History Into a Single Risk View
Rather than treating every bend, tee, and choke as equal inspection priority, the model ranks fittings by calculated erosion risk so inspection and desanding maintenance effort goes where it actually matters most.
Sand Exclusion Design
Using the Forecast to Right-Size Screens and Desanding Equipment
A sand exclusion screen or desanding cyclone sized against a single design-basis sand rate will either under-perform when actual sand production runs higher than assumed, or run needlessly conservative and expensive when it runs lower. Feeding an ongoing erosion and sand production forecast into equipment sizing decisions closes that gap on two fronts: screen slot sizing can be validated or adjusted against the actual particle size distribution the model observes over time rather than the distribution assumed at completion design, and desanding equipment duty cycle can be matched to a forecast sand load curve instead of a fixed operating assumption, reducing both wear on the desanding equipment itself and the disposal cost of handling more separated solids than the well is actually producing.
Measuring the Programme
KPIs That Show an Erosion Prediction Programme Is Working
Unplanned Choke Replacements
Target: downward trend
Count of chokes replaced due to unexpected failure versus replaced on a planned basis after risk-score flagging.
Inspection Prioritisation Accuracy
Target: risk score matches findings
How closely fittings flagged as highest-risk correlate with the fittings actually found to have the most wall loss on inspection.
Sand Cut Forecast Accuracy
Target: validated against test data
Predicted sand production rate compared against periodic sand cut test results to confirm model calibration.
Desanding Equipment Utilisation
Target: matched to actual sand load
Cyclone or separator duty cycle tracked against forecast versus actual sand load to confirm right-sizing over time.
The mistake I see most often on high sand-cut fields is treating every bend and tee on the flowline as an equal inspection priority, because that is what a fixed inspection interval forces you to do. In reality, erosion damage concentrates hard at a small number of geometry changes — usually the first bend downstream of the choke and any tee near a flow convergence point — while the straight-run sections between them barely wear at all over the same period. When you rank fittings by an actual erosion risk score instead of inspecting everything on the same calendar, the inspection budget you already have goes almost entirely to the handful of locations that were going to fail first anyway, and that is the whole game in sand management: not eliminating erosion, but making sure you always know which fitting is closest to its limit.
Tomás Rivadeneira-Osei
Production Integrity Engineer · 19 years in sand-prone onshore and offshore fields · Former asset integrity lead for high sand-cut gas fields in West Africa and the North Sea
Operations Team Questions
AI Erosion Prediction and Sand Management — Frequently Asked
Do we need acoustic sand detectors installed on every well for this to work?
No — the model is built to ingest whatever sand detection and erosion monitoring hardware a facility already has in place, most commonly a combination of acoustic sand detectors, intrusive erosion probes, and periodic production test sand cut readings, and it does not require a full new instrumentation build-out to start. Wells with only periodic sand cut test data still benefit from trend analysis and risk ranking, though the lead time on alerts is naturally shorter than for wells with continuous acoustic or probe data feeding the model. Facilities considering expanded sensor coverage can review recommended priority locations once the initial risk model is running. Book a session with our team to review your current instrumentation against the model's requirements.
How does the model account for different fitting geometries with very different erosion behaviour?
Each fitting type — elbow, tee, reducer, choke trim — carries a distinct erosion multiplier relative to straight pipe, grounded in established erosion research and CFD-validated particle tracking studies, and the model applies the appropriate multiplier for each specific fitting rather than treating the whole flowline as a single erosion risk. This is part of why risk scores can differ sharply between two fittings sitting only metres apart on the same line, and why prioritising inspection by score rather than by calendar position captures real physical differences rather than an arbitrary rotation.
Can this help us decide when to change sand exclusion screen slot size rather than just when to inspect?
Yes — one of the most direct operational uses of an ongoing erosion and sand production forecast is validating whether the screen slot sizing chosen at completion design still matches the particle size distribution the well is actually producing, since formation behaviour and sand characteristics can shift meaningfully over a well's producing life. When the observed particle distribution trends outside the range the current screen was designed for, that shows up in the model's forecast well before it shows up as a screen failure or erosion event downstream. Reach out to our team to discuss screen review workflows.
How quickly does an erosion risk score respond to a sudden change in sand production, such as after a workover?
On wells with continuous acoustic sand detection feeding the model, a sudden change in sand production is typically reflected in the risk score within the same operating day, since the model recalculates against live data rather than waiting for a periodic review cycle. On wells relying on periodic sand cut testing, the response time is bounded by how frequently that test is run, which is one of the strongest arguments for prioritising continuous monitoring on wells known to be sensitive to completion or drawdown changes.
Does this replace the need for periodic ultrasonic thickness inspection of pipeline fittings?
No — ultrasonic thickness measurement remains the ground-truth data the model calibrates its erosion predictions against, and it continues to be required for regulatory and integrity management purposes regardless of how confident the model's forecast is. What changes is which fittings get inspected most frequently and which can safely move to a longer interval, shifting a fixed inspection budget toward the locations the model identifies as highest-risk rather than spreading it evenly. Book an integrity programme review to see how this fits your current inspection interval strategy.
Sand Wears Down What You Are Not Watching Closely
Rank Your Fittings by Actual Erosion Risk Instead of Inspection Calendar Position
iFactory correlates your sand detectors, erosion probes, and choke history into a single risk view, so erosion trends surface as an alert instead of a failure report.
