Why Corrosion Prediction Models Must Be Treated with Caution
Corrosion prediction models are widely used in oil and gas projects to estimate corrosion rates, define corrosion allowance, and support integrity strategies. They are attractive because they provide numerical outputs that appear objective and reassuring.
However, corrosion prediction models are frequently misused and over-trusted, leading to underestimated risks and inappropriate integrity decisions. Many pipeline failures occur in systems where corrosion models were available, validated, and apparently conservative.
This article explains why corrosion prediction models must be treated with caution and how they should be used appropriately within pipeline integrity management.
What corrosion prediction models are designed to do
Corrosion models are developed to estimate corrosion rates under defined and simplified conditions, typically based on:
fluid composition,
temperature and pressure,
assumed flow regime,
uniform wetting conditions.
They are primarily screening and design tools, not operational integrity guarantees.
Standards such as ISO 13623 recognize corrosion prediction as an input to design and integrity planning, but not as a substitute for inspection or monitoring.
Models assume steady-state conditions
Most corrosion models assume:
steady flow,
stable chemistry,
uniform exposure of the pipe wall.
Real pipelines rarely operate under such conditions. Transients, slug flow, start-ups, shutdowns, and production changes introduce conditions that fall outside model assumptions.
As a result, model accuracy degrades precisely when integrity risk increases.
Localized corrosion is poorly captured by models
Corrosion models generally predict average corrosion rates. They do not reliably predict:
pitting depth,
under-deposit corrosion,
MIC initiation,
corrosion localization at low points.
This limitation is explicitly acknowledged in API RP 579 / ASME FFS-1, which treats corrosion models as insufficient to characterize localized damage and requires direct assessment of defects once corrosion occurs.
Input uncertainty dominates model output
Corrosion model results are highly sensitive to input assumptions:
small errors in water chemistry,
uncertainty in flow regime,
unknown distribution of free water,
variability in inhibitor effectiveness.
In practice, these parameters are often estimated rather than measured. The apparent precision of model outputs can therefore be misleading.
Models cannot anticipate operational drift
Corrosion models are typically run at design stage or early operation. They do not automatically adapt to:
gradual changes in production,
evolving water cut,
degradation of mitigation systems,
operational shortcuts or deviations.
Integrity management requires continuous reassessment, not static predictions.
Guidance such as DNV-RP-F116 emphasizes that degradation threats must be reassessed as operating conditions evolve, regardless of initial predictions.
Over-reliance on models delays corrective action
When corrosion models predict “acceptable” rates, there is a tendency to:
delay inspections,
reduce monitoring,
deprioritize mitigation improvements.
This overconfidence can persist until inspection data reveals advanced damage—often too late for low-cost intervention.
How corrosion models should be used in integrity management
Used correctly, corrosion models:
support early design decisions,
help rank relative corrosion risk,
guide initial inspection and monitoring strategies.
They must be complemented by:
field data and inspection results,
corrosion monitoring,
operational feedback,
engineering judgment.
Models support decisions, they do not replace them.
Standards perspective on corrosion prediction
Only a few standards explicitly frame the role and limitations of corrosion models:
ISO 13623 positions corrosion prediction as one element of corrosion control, to be validated and updated based on operational experience.
DNV-RP-F116 reinforces the need to reassess corrosion threats using inspection and monitoring data rather than relying on initial predictions.
API RP 579 / ASME FFS-1 addresses integrity assessment once corrosion is present, independent of predicted corrosion rates.
These documents consistently avoid treating models as integrity safeguards.
Conclusion
Corrosion prediction models are valuable tools, but dangerous when treated as truth.
Pipelines rarely fail because models were unavailable.
They fail because model predictions were trusted beyond their domain of validity.
Effective pipeline integrity management treats corrosion models as hypotheses to be tested against reality, not as guarantees of safety.