As-fabricated thickness that survives the owner's first CML sweep

Two clocks run at once: the outage window and the material lead time

A turnaround has a fixed start and a fixed end, and a fabrication shop supplying it works backwards from both. The replacement scope is set by the owner's pre-turnaround inspection data: the CMLs that came back at or below minimum thickness, the circuits with unacceptable remaining life, the dead legs and injection points flagged for renewal. That scope has to freeze early enough that material can be procured, spools fabricated, NDE completed and documentation closed before the unit comes down. Long-lead alloy, or anything requiring a mill run, pushes the freeze further out still.

Every disputed thickness reading moves the freeze. When a shop is asked on a Friday whether a spool it supplied in 2013 was Schedule 40 or Schedule 80, or what the actual wall was at the bend it formed, and the answer lives in a foreman's site file, the scope either stays open or a decision is made without evidence. Both are expensive. An open scope buys standby crews and unused staging; a decision without evidence buys a spool that either does not fit or was never needed.

This is why a shop's own thickness history is a turnaround readiness asset rather than a quality record. The question it has to answer is not whether the work was compliant — the traveller and the NDE reports already answer that — but what exactly was shipped, and whether it can be proved within the hour someone is willing to wait.

Where a fabrication shop's thickness numbers actually come from

There are four sources, and they get confused. First, the mill certificate: a declared nominal wall and a permitted tolerance, not a measurement of the length sitting in the rack. Second, receipt verification, usually a spot ultrasonic check on incoming pipe, plate or fittings, tied to a heat number. Third, in-process measurement after bending, forming or machining. Fourth, final acceptance readings on the completed component, sometimes at points the client nominates in the purchase specification.

These are different numbers with different meanings, and the field that distinguishes them is the one most shops fail to keep. A receipt reading on a random length says nothing about the specific piece that became spool 14B unless heat and piece traceability survived cutting. An as-formed reading on the extrados of a bend is the only defensible baseline for that bend and cannot be inferred from anything else, because forming loss depends on radius, method, temperature and the individual piece. A final acceptance reading at a client-nominated point is a contractual artefact and is not necessarily the thinnest point on the assembly.

A thickness history module for a shop therefore has to hold all four, each with the date, the technician and their qualification level, the instrument and its calibration status, the probe, the velocity setting and the reference block. If the module collapses them into a single thickness column, it has destroyed the one distinction that makes the data useful to anyone downstream.

Manufactured thinning is not corrosion, and the owner cannot tell the difference

Codes anticipate that forming removes wall. ASME B31.3 requires the finished bend to meet the minimum required thickness after bending, which is why fabricators start bends in heavier wall than the straight run demands. ASME Section VIII Division 1 addresses forming strain and the heat treatment it can trigger, and requires formed heads to satisfy the required thickness at the knuckle, where thinning concentrates. Nothing in either code obliges the fabricator to tell the owner where the finished component actually ended up — only that it sits above minimum.

That gap is where phantom corrosion is born. An owner's inspector places a CML on the extrados of an elbow, because that is where erosion-corrosion is expected, reads it against the nominal wall of the schedule shown on the isometric, and books the difference as service loss. If the elbow left the shop twelve percent under nominal from mill tolerance and lost another eight percent on the outer radius, the inspector has just recorded roughly a fifth of the wall as corrosion that never occurred. The consequence is not academic: corrosion rate is a divisor in a remaining-life calculation, so an inflated rate shortens the next inspection interval and can move a perfectly serviceable line onto a replacement list in the following turnaround.

The reverse error happens too. A shop that supplies heavier wall than specified, substituting a heavier schedule because that is what was in the rack, creates a component that reads thick for years and then appears to corrode alarmingly fast when someone finally trends it against the correct as-built figure. Recording what was actually shipped, piece by piece, is the only cure for either direction of the error.

The handover record is the owner's baseline, whether you write it or not

When a replacement spool goes into a circuit, the owner's inspection programme has to begin monitoring it. Following API 570 practice, the inspector establishes or re-establishes CMLs, and the first reading taken becomes the baseline by default. If the shop supplied an as-built datasheet, that baseline is a measurement made on clean, uncoated, accessible steel by a qualified technician with a calibrated gauge. If it did not, the baseline is whatever gets read through insulation cladding on a hot line six months into service, by a crew working from a scaffold in a rainstorm.

A shop that ships the as-built CML datasheet with the spool — location on the isometric, weld number, heat, as-formed minima at the bend extrados and head knuckle, method, technician, instrument, date, and every point that could not be read with the reason — is doing something no competitor's quotation mentions and every integrity engineer notices. It also protects the shop. When a low reading appears two years later, the argument about whether it was shipped that way is settled by a document rather than by two people's recollections.

The same record feeds the owner's management of change. Where a replacement is not like-for-like — a different schedule, a different material, or a re-rating under API 510 provisions — the documented as-built thickness forms part of the evidence that the change was properly evaluated, and OSHA's process safety management requirements make that evidence auditable rather than optional.

Points that cannot be read, and the discipline of saying so

Fabricated geometry defeats ultrasonic thickness measurement in entirely predictable places: the inside radius of a short-radius elbow, the crown of a weld cap, socket weld fittings, threaded connections, branch reinforcement pads, backing rings, cladding or overlay on the bore, galvanised or freshly painted surfaces, and anything already insulated before the reading was scheduled. A crew that cannot obtain a reading does one of three things. It omits the row, it substitutes a reading from somewhere nearby, or it writes a note that nobody structures.

Silent omission is the worst, because it is invisible in every subsequent report. Undocumented substitution is worse still in the long run, because a reading taken 150 mm away on the straight section will look like a valid extrados value for the life of the asset. The discipline that costs nothing and repays repeatedly is to create the record for the point, mark it as not taken, name the reason from a controlled list, and state what would recover it: surface preparation, cladding removal, a different probe, or a radiographic profile shot for a geometry ultrasonics cannot address.

In turnaround planning this list is directly actionable. Points that cannot be read on a spool in the shop are usually the points that cannot be read in the field either. Knowing that eight weeks before the window means scaffolding, insulation removal or an alternative technique can be scheduled deliberately, rather than discovered by a crew standing in front of the line on day three.

Qualification and calibration on every line, because the audit lands mid-window

Shop thickness readings are taken under a written practice — ASNT SNT-TC-1A or CP-189, or ISO 9712 where the client specifies it — by personnel qualified for ultrasonic thickness measurement at the appropriate level. The gauge carries a calibration due date. The reference block has an identity. The velocity setting has to match the material. Each of these is trivially available on the day the reading is taken and effectively impossible to reconstruct a year later.

Turnaround work is precisely where reconstruction gets demanded. Client inspectors and third-party auditors review supplier records during the window rather than before it, and the finding that holds a spool is rarely the reading itself. It is a technician whose certification lapsed between fabrication and delivery, or a gauge whose calibration expired on a date falling between two readings on the same sheet. A module that binds qualification validity and calibration status to the reading at the moment it is recorded makes that finding impossible to generate.

The same binding makes the shop's own internal review worth doing. Filtering every reading taken on an instrument that was out of calibration during the last quarter takes seconds, and either returns nothing — which is the answer you want before an audit — or returns a bounded list that can be re-taken quietly, on your own schedule, before anyone external asks the question.

Evaluating a module against a real turnaround

Ask for the module to be demonstrated against your own last outage rather than a demonstration dataset. For one spool, can it produce every thickness value ever recorded against it, with date, technician, instrument, method and the type of reading? Can it separate receipt, as-formed and final acceptance values? Does it maintain heat and piece linkage through cutting? Can it record a not-taken point with a controlled reason and report those across an order? Can it emit a handover datasheet in a structured format the client's inspection system can ingest, instead of a PDF somebody retypes?

Then ask the schedule questions, which are the ones a project engineer actually lives with. Can they see, for the current turnaround package, which spools are missing final readings, which carry readings from an instrument since found out of calibration, and which have unresolved not-taken points, in a single view? That view is the difference between finding a documentation gap in week three of fabrication, when it costs an afternoon, and finding it during the window, when it costs the window.

Atlantis configures this on an Odoo-based inspection management platform shaped around fabrication routing, spool numbering and turnaround packages rather than a generic work order. It is affordable, accessible and fully customisable to a shop's own drawing conventions. To see it run against one of your own spool packages, request a demonstration at info@atlantisndt.com.

Request a consultation