What changes when CML history lives on the geometry, not a spreadsheet

Four things change. The inspector finds the CML on the model instead of hunting a drawing index. Corrosion appears as a spatial pattern instead of a column of numbers. An incoming contractor inherits context rather than a folder. The next campaign is planned from where the wall is thinnest. What breaks it is the maintenance of the CML-to-object link.

The change is not visual. A CML is a designated area containing one or more examination points, and both API 510 and API 570 expect it to be marked on inspection drawings so that every campaign measures the same place — because the corrosion rate arithmetic subtracts one reading from another and attributes the whole difference to metal loss. When the location moves, the arithmetic still runs and reports a rate that is really a location difference. Putting the CML on the geometry fixes that repeatability assumption directly rather than decoratively. Three further changes follow: corrosion becomes a spatial pattern with a mechanism attached instead of a column sorted by an identifier nobody assigned for physical reasons, a new contractor inherits context rather than a folder of PDFs, and campaign scope is ranked by remaining life rather than by due date. What decides whether any of it survives is maintenance of the link between CML identity and model object, and the thickness trending view is where that link proves itself.

Source: API 510 and API 570 terminology for condition monitoring locations, which replaced the earlier thickness measurement location designation; a CML is a designated area containing one or more examination points, expected to be marked on inspection drawings so measurements repeat at the same location. Corrosion rate and remaining life relationships as applied under API 510 and API 570: long-term rate from initial and current thickness, short-term rate from previous and current thickness, governing rate the higher of the two, remaining life from current thickness less required thickness divided by the governing rate. No clause numbers cited; requirements described against the codes' current editions as published by API. Verified August 2026.

Technically reviewed by Anoop Rayavarapu — ASNT NDT Level III (UT, RT, MT, PT, VT)
Six inspection tasks, before and after the data sits on the geometry
TaskRegister and drawing indexOn the twinWhat actually changes
Locating a CML in the fieldRead the ID, find the isometric, count welds or fittings from a reference pointNavigate to the object, confirm against surrounding geometry and a site photographReadings repeat at the same square inch, which is what the corrosion rate formula assumes
Reading a corrosion patternA column sorted by CML ID, an order that carries no physical meaningThickness rendered onto the geometry itselfElbow extrados, six o'clock runs and injection-point bands become visible as mechanisms
Finding monitoring gapsThe absence of a CML is invisible in a tableUnmonitored branches, dead legs and small-bore connections show as blank regionsCoverage gaps get found by review rather than by a leak
Onboarding a new inspection contractorA folder of PDFs plus one or two campaigns of learning the unitAn access grant to the unit's full history, in placeFirst-campaign quality stops being measurably worse than the fifth
Planning the next campaignA work list assembled from due datesThe unit ranked by remaining life and damage activityHours move from calendar-due points to genuinely thin points
Handling a replaced spoolThe new component silently inherits the old thickness historyOld CML closed, new CML opened with a new baselineFabrication differences stop being reported as corrosion rates
None of this improves the measurement itself. Reading quality remains governed by surface preparation, couplant, probe selection, temperature compensation, calibration and technician qualification.

A CML is a location before it is a number

API 510 and API 570 replaced the older thickness measurement location designation with the condition monitoring location. A CML is a designated area on the equipment where periodic examinations are performed to assess and monitor condition, and it contains one or more examination points — commonly a marked circle a few inches across on the surface. The codes expect CMLs to be marked on inspection drawings so that the same place is measured every campaign, because the entire corrosion rate calculation rests on repeatability of location.

That assumption is where spreadsheet-based programmes fail quietly. The formulas are unforgiving about it. The long-term rate is the initial thickness minus the current thickness divided by the years between them. The short-term rate is the previous thickness minus the current thickness divided by the years between those. The governing rate is the higher of the two. Remaining life is the current thickness minus the required thickness divided by that rate. Every term in both formulas assumes the two readings came from the same square inch of steel.

When a technician measures 40 mm from the weld instead of 200 mm because the drawing was ambiguous, the arithmetic still runs. It produces a corrosion rate that is a location difference wearing the costume of a metal loss trend, and nothing in the spreadsheet flags it. Attaching the CML to geometry is not a visualisation upgrade. It is a fix for the repeatability assumption that the codes require and that paper isometrics enforce weakly.

Change one: finding the CML

The current workflow is a drawing index. The inspector opens the register, reads a CML identifier, finds the isometric that identifier belongs to, locates the balloon on the isometric, walks to the line, counts welds or elbows from a reference point, and marks up. In a congested unit with parallel lines in the same service, that last step is the one that goes wrong, and it goes wrong most often with a contractor crew who did not set the CMLs originally and never met the person who did.

On a twin the CML is an object in space. The inspector navigates to it, sees it in context against the surrounding geometry — which elbow, which side, how far from which flange — and confirms the position visually before climbing. On a tablet at the job site that comparison takes seconds against a photograph of the actual pipe. The number of readings taken at the wrong location falls because the ambiguity that produced them is gone, not because anybody tried harder or was told to be careful.

The knock-on effect is in scaffolding and access. When the CMLs for a campaign can be selected spatially, the access plan is built from where the points are rather than from a list of line numbers, and points sharing a scaffold get grouped into one visit. That is a scheduling gain which lands in the shutdown budget rather than the inspection budget, which is why it is usually the number that justifies the programme to the people who approve it.

Change two: corrosion as a pattern, not a column

A thickness register is a column of numbers sorted by CML identifier, and identifiers are assigned in the order somebody walked the line. The sort order carries no physical meaning, so a pattern distributed in space is invisible in the table. Four readings thinning faster than their neighbours look like scatter in a column. Plotted on the geometry, the same four readings sit on the extrados of consecutive elbows, or at the six o'clock position along one run, or immediately downstream of an injection point.

Those three patterns are three different damage mechanisms with three different responses. Extrados thinning through elbows is erosion-corrosion driven by velocity and geometry. Six o'clock thinning along a run is under-deposit or wet-phase attack. A thinning band downstream of an injection point is exactly where API guidance says to concentrate monitoring. Each one changes where the next campaign looks and what the engineer does about it, and none of them is legible in a list sorted by an identifier.

The corollary is that spatial display exposes gaps as clearly as it exposes damage. A heat map with a dark region is not always good news — it is often a region with no CMLs at all, and a table sorted by identifier never shows an absence. Programmes that move onto a twin routinely find whole branches, dead legs and small-bore connections with no monitoring at all, which is a finding worth more than the visualisation that produced it. The thickness trending view is where that reads clearest.

Change three: handover context instead of a folder

Inspection contracts change hands, and today the handover is a folder — a data dump of PDF reports, a thickness spreadsheet, a set of isometrics of uncertain vintage, and the tacit knowledge of the outgoing inspector, which does not transfer at all. The incoming crew rebuilds an understanding of the unit over one or two campaigns, and the first of those campaigns is measurably worse: more repeat visits, more located-the-wrong-point corrections, more questions escalated to the owner's engineer at the worst moment.

When the history lives on the geometry, the handover is an access grant. The incoming inspector opens the unit, selects a line, and sees every reading ever taken on it — when, by whom, with what instrument, against which report — in place and in context. What previously took two campaigns to learn is available on the first morning. The value is largest exactly where turnover is highest, which is contract inspection, and it is the argument that gets the programme funded in organisations that outsource inspection entirely.

The same structure serves an audit. When a regulator or a client asks why an interval was extended, the answer is the geometry with the data on it, the readings that supported the decision, and the qualifications of the people who took them. That is the same evidence chain that makes an NDT report defensible, assembled once and kept, rather than reconstructed from four systems under time pressure while an auditor waits.

Change four: planning from the thin end

Campaign scope is normally built from intervals — this vessel is due, that circuit is due — and the resulting work list is a calendar artefact. Remaining life is calculated per CML and sits in the register, so the unit's thinnest points are already known and scattered across it. Rendered on the geometry, the same data ranks the unit by where the wall is actually thinnest, and the campaign is built from the top of that list downward until the budget runs out.

That reordering changes what gets inspected. Points with decades of remaining life stop consuming inspection hours on a calendar cadence, and points approaching their limit get earlier and denser coverage. It is the input side of a risk-based programme, and a spatially resolved thickness picture is the most credible corrosion-likelihood evidence an RBI programme can be built from — considerably better than a generic circuit corrosion rate applied uniformly to everything in the same service.

It also changes what a finding triggers. When a reading drops below the required thickness, the surrounding CMLs are immediately visible as neighbours rather than as identifiers to look up, so the extent of the affected region is scoped in minutes instead of an afternoon. If a fitness-for-service assessment follows, the assessor starts with the geometry and the local thickness field already assembled, rather than reconstructing both from a spreadsheet and an isometric of unclear vintage.

The maintenance burden that makes or breaks it

Everything above depends on one link staying true: the CML identifier in the inspection database pointing at the correct object in the model. That link has no natural custodian and four reliable ways to break. A database migration renumbers CMLs while the model keeps the old identifiers. A spool is replaced and the new component inherits the old thickness history, manufacturing a corrosion rate out of a fabrication difference. A management-of-change modifies the plant and the geometry is never updated. Readings arrive with identifiers matching nothing and land in an orphan pile.

The governance that holds it together is small and boring. One system owns CML identity and everything else references it — the inspection data management side of the ERP is the natural home, because that is where the register already lives and where the calculations already run. Any management-of-change that touches geometry raises a model update task in the same workflow that raises the drawing update. An orphan report runs after every campaign with a named owner and a deadline. Component replacement closes the old CML and opens a new one with a new baseline.

Programmes that skip this get eighteen good months and then a twin nobody trusts, which is worse than the spreadsheet because it looks authoritative. The failure is never dramatic. Someone notices a spool on the model that was cut out last turnaround, and from that day the whole model is treated as indicative. Budget the maintenance as an operating line with a named owner before the twin is built, not as a project closeout task assigned to whoever is left.

What it does not change

A twin does not make a reading correct. A UT thickness reading placed perfectly on a model is still governed by surface preparation, couplant, probe selection, temperature compensation, calibration against a block of the right material and velocity, and the qualification of the technician who took it. Placing a wrong number accurately produces a confidently wrong corrosion rate, and the spatial display makes it more persuasive rather than less. The measurement quality problem sits upstream and it stays upstream.

That is why the data-quality layer belongs alongside the twin rather than after it. Reading-level validation catches what a spreadsheet accepts silently: instrument calibration in date, technique sheet followed, technician certified to the method and level, values inside physically plausible bounds against the previous reading at the same point. Independent report review and Level III oversight do the same job for the reports those readings arrive in, before the numbers reach a trend line.

It also does not change the code. Intervals, required thickness calculations, corrosion rate selection and remaining life all run exactly as they did, on the same formulas, subject to the same engineering judgement and the same authorised inspector. What the twin changes is how quickly the right data reaches the person applying them, and how confident everyone is that the readings came from where the record says they did.

A rollout that survives the first turnaround

Start with one unit and the CML register you already have. No re-survey, no new numbering scheme, no attempt to model the whole site. The geometry can come from an existing scan, and whether that scan is reusable is settled on the point cloud formats page rather than by asking the original contractor. Load the existing register, link what links, and publish the orphan list. The orphan list is the first deliverable and the first honest measure of the inspection data you actually hold.

Prove it in one turnaround. Run the campaign with the twin alongside the existing process rather than replacing it, and count three things: readings taken at a location the inspector had to query, scaffold visits that served more than one CML, and hours spent by the incoming crew locating points. Those three numbers are the business case, and all three are measurable inside a single outage rather than inferred from a vendor's benefits slide.

Then extend by service severity rather than by unit boundary. The circuits with the shortest remaining life and the most active damage mechanisms earn the geometry first; the tank farm waits. Atlantis builds this on the Digital Twins platform against your existing inspection register — affordable, accessible, and fully customisable to the data you already keep. Request a demo with one unit's register and one existing scan, and the orphan list comes back before anything is committed.

What actually changes when inspection data sits on a 3D model?

Four things. Locating a CML becomes navigation instead of counting welds off an isometric. Corrosion becomes a spatial pattern with a mechanism attached instead of a sorted column. A new contractor inherits context on day one instead of learning the unit over two campaigns. And campaign scope gets ranked by remaining life rather than assembled from due dates.

Do I need a new CML numbering scheme to build a twin?

No, and creating one is the fastest way to fail. Use the register you already have. The build links existing CML identifiers to model objects and publishes an orphan list of the ones that do not resolve. That orphan list is the first real deliverable, because it measures the quality of the inspection data you actually hold rather than the data you assume you hold.

What breaks a digital twin of an inspection programme?

The link between CML identity and model object, in four reliable ways. A database migration renumbers CMLs while the model keeps the old identifiers. A replaced spool inherits the old thickness history. A management-of-change alters the plant and the geometry is not updated. Readings arrive with identifiers matching nothing and accumulate in an orphan pile with no owner and no deadline.

Does a digital twin replace my inspection data management system?

No, and it should not try. The twin is the spatial index onto that record. One system owns identity and everything else references it — that separation is what keeps the twin honest when the register is migrated, extended or audited.

Can a twin be built from a point cloud we already have?

Yes, if the existing survey preserved per-setup structure and is readable outside the contractor's software. That is a file format question rather than a scanning one, and it is settled by checking whether an open master exists alongside the indexed working copy. Where the old cloud is per-setup and undecimated, no re-survey is needed to start on one unit.

Does this help with an API 570 or API 510 audit?

It shortens the answer. When an auditor asks why an interval was extended or why a CML was retired, the evidence is the geometry, the readings on it, the dates, the instruments and the qualifications of the people who took them — assembled once rather than reconstructed under time pressure. The codes and the calculations are unchanged; the retrieval time collapses.

Request a consultation