Corrosion Rate Predictions That Survive Bad Readings

In asset integrity management, corrosion-rate predictions drive decisions that matter: predicted wall thickness, Remnant Life (RL), and the Next Inspection Date (NID). Ultimately, these predictions help determine whether equipment is scheduled for re-inspection next year or five years from now.

The challenge is that a predicted corrosion rate is only as trustworthy as the data behind it – and field wall-thickness (WT) data is rarely clean. Corrosion, inspection, and reliability engineers regularly encounter measurement outliers, undocumented repairs, and changes in process conditions. If these events are not recognized and interpreted correctly, the predicted corrosion rate may no longer reflect actual conditions in the field

Consider an inspection engineer reviewing wall-thickness measurements after a routine wall thickness survey. A new reading from a corrosion monitoring location (CML) is significantly different from what the circuit’s history would suggest. As a result, the projected corrosion rate rises sharply, reducing the calculated RL and bringing the NID forward.

There are two broad explanations: either the corrosion behavior has genuinely changed, or the reading is unreliable. A prediction based on the wrong explanation cannot be trusted.

In IMS, the Short Corrosion Rate (Short CR) compares the previous and latest readings at a CML and uses the result to calculate RL and NID. It is a fast and useful check, but it considers only two data points. An error of just one or two millimeters at a single CML can dramatically increase (or decrease) the calculated Short CR, and the calculation cannot determine on its own whether the change is real.

The other conventional methods face similar limitations:

Although these methods use the data differently, none can independently distinguish a bad reading from a genuine change in wall-thickness level or corrosion rate before producing a prediction.

In practice, wall-thickness measurements may be taken by different crews, using different equipment and at irregular intervals. Scatter and outliers are therefore common rather than exceptional.

The consequence is more than an incorrect number in a report. An overestimated CR can place a circuit in a shorter Remnant Life category, triggering an unnecessarily early inspection. A prediction that misses a genuine acceleration in corrosion creates the opposite risk: the change may go undetected until a piece of equipment fails.

Either way, engineers should not have to manually reanalyze years of data simply to determine whether a corrosion rate prediction is credible.

The Integrity Data Analysis and Plotting (IDAP) module in Cenosco’s IMS (Integrity Management System) is designed to address this problem. Rather than committing to a single explanation as soon as a new reading arrives, IDAP evaluates several possible explanations and uses the evidence to determine how the forecast should respond.

IDAP uses a multi-process Dynamic Linear Model (DLM). This forecasting method continuously evaluates how well four competing models explain the incoming data:

Returning to the earlier example, this analysis helps prevent a surprising reading from automatically producing a misleading forecast. When a new wall thickness reading arrives, IDAP evaluates it against all four models. Is it part of the established trend, a measurement outlier, a step change following an undocumented repair, or evidence of a genuine change in corrosion rate?

When Model 2, 3, or 4 is more strongly supported than the routine Model 1, IDAP raises an anomaly flag for review. It can then account for the nature of the event instead of allowing the new reading to influence the forecast without context. An isolated outlier can be prevented from distorting the trend, while a genuine wall thickness level or corrosion rate change can be recognized appropriately.

This is what distinguishes IDAP CR from Long CR, Short CR, and Linear Regression CR: IDAP assesses how each reading should influence the prediction rather than assuming that every reading represents the same underlying trend.

A trustworthy corrosion-rate prediction is not necessarily built on the most data. It is built on the correct interpretation of the data.

IMS makes that interpretation part of a focused review rather than a manual reanalysis. Engineers can inspect anomalous readings, review the corresponding trend plots, confirm how flagged data should be handled, and approve the circuit’s CR calculation together with the resulting RL and NID.

This provides both analytical support and engineering oversight. IDAP identifies the readings that require attention, while the engineer retains control over the final assessment.

So far, we have considered one anomalous reading on one circuit. During a unit turnaround or periodic integrity review, however, engineers may need to generate trustworthy corrosion rate predictions for dozens of circuits.

Bulk Analysis displays overview plots for all selected circuits on a single screen. Reliability engineers and Technical Authorities can screen an entire unit for anomalies, then use single-circuit Normal Analysis to investigate only the circuits that require closer review.

This helps teams focus their attention where engineering judgment adds the most value, without manually re-analyzing every historical reading for every circuit.

A corrosion rate prediction is only as reliable as its ability to distinguish a genuine trend from a bad reading before forecasting forward. IDAP builds that distinction into every forecast, one reading and one circuit at a time.