Tracking equipment degradation between inspections: method and conditions

One inspection gives you a snapshot. Two inspections give you a direction. Three give you a rate. It is that rate which lets you decide when to come back, and it is that rate which is almost always missing. Not because measurements are lacking, but because those measurements have never been brought alongside one another.

What a single reading does not tell you

A wall thickness recorded at 8.2 mm means nothing on its own. It only takes on meaning when you compare it with the original nominal thickness, with the minimum allowable thickness for the service conditions, and above all with the values previously recorded at the same spot.

It is this last comparison that causes trouble in practice. Working out the loss against the nominal figure is immediate. Comparing over time means retrieving the earlier campaigns, identifying the same condition monitoring location, and making sure it really is the same one. Across a fleet of several dozen items followed for ten years by a succession of contractors, that work is rarely done except on the most closely watched equipment.

The result is familiar: you hold a great many measurements and very few trends.

The decisive condition: a stable measurement location

A condition monitoring location, or CML, is a precise physical point on a piece of equipment. For two readings to be comparable, they must fall on the same spot, to within a few millimetres. That means the location has to be identified in a durable way (a physical mark, a location sketch, a reference photograph) and that this reference has to be carried over from one campaign to the next.

This last point deserves some emphasis. Keeping "minimum thickness recorded: 7.9 mm" instead of the point-by-point readings is enough for a report, but it makes any real tracking impossible. That minimum value may well come from a different location at every campaign, so comparing one campaign's minimum with the next tells you nothing about how any single point is actually behaving.

Reading a trend without being misled

Once the readings have been brought together, interpretation calls for caution.

A variation is not always degradation. Measurement uncertainty is real. A difference of a few hundredths of a millimetre between two campaigns may reflect the repeatability of the measurement rather than any loss of metal. You need to know the order of magnitude of that uncertainty before drawing a conclusion. Treat any change smaller than the uncertainty band as noise until several campaigns line up in the same direction.

A rising value signals a method problem. Wall thickness does not grow back. If a reading is higher than the previous one by more than the uncertainty, the explanation lies elsewhere: a different location, a different technique, a data-entry error or a mis-linked point. That is a signal to investigate, not a data point to smooth away.

A trend built on two points is fragile. Two values always define a straight line. You need at least three campaigns before you can begin to tell steady degradation apart from a measurement mishap.

An acceleration matters more than a value. A location that loses the same amount every year is predictable. A location whose loss is speeding up signals a change of regime (a change of service, localised degradation, a failing coating) and justifies a closer look well before the calculated due date.

What putting readings into a trend lets you decide

Degradation tracking is not about producing curves. It is about settling four questions.

• When to come back? The interval can be tuned to the observed rate rather than applied uniformly across the whole fleet.
• Where to concentrate the effort? The points that are moving deserve more attention than those that have been stable for ten years.
• What to prepare? A due date worked out in advance lets you order a part or slot an intervention into a turnaround rather than deal with it under pressure.
• What to tell an auditor? A documented trend, linked back to its source reports, is a solid answer on how well the equipment is being kept under control.

Estimating the remaining useful life follows directly from this work, with its own precautions: see estimating remaining useful life. Ranking the effort according to criticality is covered in prioritising maintenance by real risk.

Common mistakes

• Comparing minima instead of comparing pointsone campaign's minimum may come from a different location than the previous campaign's. The comparison then means nothing.
• Smoothing away outliersan inconsistent value is information about the method. Deleting it removes the only signal available.
• Extrapolating from two pointstwo readings are not enough to establish a rate. Saying so plainly is better than displaying a reassuring due date.
• Ignoring a change of techniqueswitching from one measurement technique to another introduces a systematic offset that has to be documented, otherwise it will be read as degradation.
• Keeping only the report summarythis is the most frequent cause of unusable histories. The raw values must be retained.

Making the tracking sustainable over time

Degradation tracking rarely fails for technical reasons. It fails because it rests on one person keeping a spreadsheet, and it stops the day that person moves on.

Three decisions make it durable. First, fix the equipment and CML references as reference data, not as free-text labels. Second, systematically keep the source report alongside the value, so that any anomaly can be arbitrated. Third, feed the tracking as each report arrives rather than through an annual catch-up. See centralising inspection reports.

This is the logic any integrity tool should follow: rebuild the history by equipment and by condition monitoring location from the existing reports, keeping the link to the source document. On the equipment families concerned, see in particular piping and pressure equipment, or the Maintenance Intelligence page for the overall approach.