Tracking the condition of your industrial equipment: asset integrity management

Asset integrity management is about knowing, at any given moment and for every piece of equipment, what condition it is in, what has been observed on it, when it next needs looking at, and what happens if nothing is done. Inspection managers have always carried out this work by hand, with ring binders and spreadsheets. What has changed today is the ability to hold that picture together without spending weeks a year on it.

What the term actually covers

The expression asset integrity describes the ability of a piece of equipment to perform its function, safely, throughout its whole operating life. It applies first and foremost to equipment whose failure carries serious consequences: pressure vessels, piping, heat exchangers, steam generators, load-bearing structures, and storage tanks.

Managing that integrity is therefore not the same as managing maintenance in the broad sense. Replacing a fan bearing is a maintenance job; tracking the remaining wall thickness of a storage-tank shell course over fifteen years is an integrity matter. The two meet at the point where an intervention has to be decided, but they do not rest on the same data or on the same time horizons. Maintenance thinks in shifts and weeks; integrity thinks in campaigns and decades, and the difference in tempo is exactly why the two disciplines need different records.

An integrity system is organised around the equipment, not around the intervention. That is a reversal of perspective, and it has very concrete consequences for how information is filed. In a work-centred system, the natural unit is the job order and everything hangs off it. In an equipment-centred system, the natural unit is the asset itself, and every report, every measurement, every deadline is attached to it and stays attached to it for the whole of its life.

The four building blocks of an integrity system

The equipment register and its criticality

Everything starts with a list. Not a list of machines, but a list of tracked objects: a tag number, a family, a site, a function, and its service characteristics. Without a stable tag, nothing else can be hung on it. The tag is the spine of the whole system, and a register that is vague about identity will stay vague forever.

Onto that register a notion of criticality is added: what the consequences of a failure would be, in terms of safety, of production, and of compliance. This is what later makes it possible to rank the effort, because no site has the resources to watch everything with the same intensity. We set out how that ranking is built in the article on the criticality matrix.

The history of observations

This is the heart of the matter, and it is almost always the weak point. The history is every inspection report, every non-destructive testing record, every requalification, and every statutory visit produced since commissioning, most often by several different contractors, in different formats, and over different scopes. Twenty years of a plant's life can sit in a dozen incompatible shapes.

As long as that history stays in the form of documents, it is consultable but not usable. You can find a report; you cannot answer, in thirty seconds, "which items of equipment have had an open defect for more than two years". That gap is the subject of the article on how to centralise inspection reports. The point is not that the documents are missing; it is that the information locked inside them cannot be queried.

The measurements and their trend

The wall thickness readings taken by ultrasonic testing, the clearances, the temperatures, the vibration levels: these values only mean something when they are compared with themselves over time. A single thickness reading tells you nothing. Three thickness readings at the same condition monitoring location on three different dates draw out a corrosion rate, and therefore a deadline. That is the whole reason the readings are worth keeping.

This is why an integrity system imposes a discipline on condition monitoring locations: a CML must keep the same identity from one campaign to the next, whatever contractor took the reading. Move the point by a few centimetres, or renumber it between campaigns, and the trend that was supposed to protect you quietly becomes meaningless. See tracking the degradation of a piece of equipment.

The deadlines and the obligations

Requalifications, periodic visits, and the checks imposed by the site's quality framework: these dates are not negotiable. An integrity system keeps them up to date and flags what is coming, with the lead time needed to prepare for it, rather than the day before. A deadline that surfaces late is not a scheduling detail; on a pressure vessel it can mean an unplanned shutdown.

What separates an integrity system from a CMMS

The confusion is common, and it becomes expensive when it leads to buying the wrong tool. Both are useful, but they answer different questions.

Table scrolls horizontally on small screens.

A CMMS organises the work. An integrity system organises the condition. We develop this comparison, together with the selection criteria, in how to choose a CMMS.

The right configuration, on the great majority of sites, is to keep the CMMS in place and bolt on the missing building block rather than replace everything. Replacing a CMMS that is already in service opens up a change-management project that has nothing to do with the original need, and it usually costs far more time than it saves.

Where to begin when everything is still inside PDFs

What AI does, and what it does not do

The automatic analysis of reports changes one specific thing: history migration, which until now has been a deal-breaking obstacle. Reading two hundred reports by hand to extract observations and measurements represents an amount of work that nobody funds. Having a machine read them makes the operation feasible, and it turns a paper archive that was merely stored into a record that can be queried.

You have to be clear about the limits, because they determine the credibility of the whole thing:

• AI does not pronounce a fitness-for-service verdictit puts no signature on the line, decides no requalification, and replaces neither the inspector, nor the accredited body, nor the manager who owns the decision.
• It does not fill in missing dataif a wall thickness was never taken at a location, no amount of processing will make it appear. A serious tool flags the gap instead of offering a plausible value.
• It does not make an illegible document reliablea poor-quality scan must produce a flag, not an estimate. The rule to demand of a supplier is that the tool says "I cannot read this" rather than producing an approximate value.
• It does not remove the need for traceabilityevery value displayed must be traceable back to the document and the page it came from. A tool that does not keep its sources puts you in difficulty at audit instead of helping you.

These limits are not statements of principle: they are the very points on which an auditor or an operator will question you. A system that does not hold to them will serve you only once. We set out this division of roles in why human validation remains indispensable.

How to tell whether you need one

A few questions are enough to settle it. Put them to your team, with a stopwatch in hand.

• How long does it take to say which items of equipment have a deadline falling within the next six months?
• Can you show the change in wall thickness of one specific item without reopening the PDFs one by one?
• If the person who keeps the thickness spreadsheet leaves tomorrow, what is left?
• At the last audit, how many days went into pulling the evidence together?
• Do you know, today, how many observations are open and for how long?

If three of those five answers are uncomfortable, the subject is not maintenance: it is integrity. And the current cost is not zero: it is simply spread across invisible hours, scattered through the year, that appear in no budget line.

What we have taken from it

Tools capable of reading inspection reports in PDF form and reconstructing the history per item of equipment (observations, changes in measurements, deadlines) exist today. Watching them work on real plants teaches three things.

The first: the obstacle is almost never technical, it is documentary. The data exists, it is simply locked away, and unlocking it is a filing problem far more than an engineering one.

The second: fidelity beats cleverness. A tool that gives back exactly what the report says, including its odd wordings, is worth more than a tool that interprets and gets it wrong one time in twenty. In integrity work, the wrong value confidently reported is more dangerous than no value at all.

The third: local processing is not a detail. Many industrial operators refuse to let the real condition of their installations leave the site network, and that requirement is dealt with at the architecture stage, not afterwards.

To go further on the implementation side of maintenance, see our Maintenance Intelligence page; for audit preparation, see Quality & Compliance.