Standardising Deficiency and Recommendation Tracking Across Multiple Steel Plants

Severity is where standardisation actually breaks

Groups that set out to standardise deficiency tracking almost always start with the form. They agree a common finding template, push it to every mill, and discover a quarter later that the corporate dashboard still cannot be trusted. The form was never the problem. Site A grades a wall-loss indication on a blast furnace gas main as Category 2 on a scale that runs 1 to 4 with 1 as most severe. Site B grades the identical indication as a 3 on a 1-to-5 scale where 5 is most severe. Both entries are locally correct. Averaged together they are meaningless, and a chart built on them is worse than no chart because it carries the authority of a number.

The correct treatment is not a mandated rename. It is a severity model that stores the site's own grade verbatim alongside a derived corporate grade, with the mapping held as versioned data that a named integrity engineer at each mill signs off. When a site revises its scale, you change the mapping, not the history. A module that overwrites the local grade during import has destroyed the only thing that lets you re-derive a figure when a plant manager challenges it in a group review, and that challenge will come the first time the number is used to allocate capital.

Ask any vendor to show two records from different sites resolving to the same corporate severity, then show the mapping that got them there and who approved it. If the answer is that findings are standardised on import, the roll-up is a one-way transformation and the group has bought a number it can never explain back to the plant that produced it.

A mill is not a plant with a P&ID

Process industries have an easier taxonomy problem than steel because a refinery unit has a line list and a circuit definition that everyone accepts. A steel plant does not. The same physical crane is the ladle crane to the melt shop, crane 4 to maintenance, and asset 112-CR-04 in the fixed asset register. Findings raised against the three names never join up, and the group's most common asset failure mode disappears into three small piles.

ISO 14224:2016 supplies a hierarchy that works once adapted: installation, plant, section or system, equipment unit, subunit, component, down to maintainable item. The specific labels matter less than every site placing findings at a consistent level. If one mill tags a crack at line level — hot strip mill — and another tags it at component level — bearing housing, drive side, stand 6 — no cross-plant failure-mode analysis is possible, and the group cannot answer whether roll neck cracking is a design issue or a lubrication issue.

Practically, this means the deficiency module needs functional-location import from each site's existing register with alias handling, not a blank hierarchy someone has to populate. Expect the first pass to reveal duplicates, orphaned locations and assets that were decommissioned years ago but still receive findings. That clean-up is the actual standardisation work, and a module that cannot show which locations have received a finding in the last three years makes it far harder.

What closed has to mean when the evidence is a re-shoot

The single largest source of incomparable open-deficiency counts is the close criterion. One mill closes a finding when a work order is raised, because from its perspective the inspection function has done its job and handed over. Another closes only after a passing re-inspection. The first site's open count will always look better, and it is the site carrying more unverified risk.

Enforce evidence-based closure and the counts become comparable overnight. For a linear indication on a ladle crane trunnion weld, evidence means the re-inspection record: method and technique, the operator's certification level under SNT-TC-1A or CP-189, the acceptance criteria applied, the equipment and its calibration status, and the date. For a wall-loss finding on a B31.1 header, evidence means the follow-up thickness set at the same grid locations, not a general statement that a spool was replaced. OSHA 29 CFR 1910.179(j) already obliges a dated certification record for crane inspections, so for a large part of the mill the evidence exists — it is simply not attached to the finding.

There is a second-order effect worth planning for. The first quarter after evidence-based closure goes live, the group open count rises sharply, because findings that were administratively closed re-enter the queue. Brief the executive sponsor before the switch, not after. A module that lets you run both definitions in parallel for one reporting period turns that spike from a crisis into a footnote.

Cranes, ladles and the records that are not discretionary

Mobile and overhead lifting is where a steel group's deficiency register meets a federal inspection requirement rather than an internal standard. OSHA 29 CFR 1910.179 sets frequent and periodic inspection intervals for overhead and gantry cranes and requires records of periodic inspections including the date, the examiner, and identification of the equipment. ASME B30.2 and B30.17 supply the removal-from-service criteria — hook throat opening, twist, wire rope broken-wire counts per lay, chain elongation. CMAA Specification 70 governs the structural side for top-running bridge cranes.

In a hot metal environment these are not paperwork items. Ladle crane structural failures are the classic catastrophic event in a melt shop, and the load path — trunnion, hook block, main girder connections — is exactly where MT and UT findings accumulate. The deficiency module has to treat a crane finding as a first-class record with the same evidence and ageing discipline as a pressure boundary finding, not shunt it into a maintenance checklist that never reaches the integrity report.

A specific trap: crane inspection findings are frequently recorded on paper at height and keyed in days or weeks later. If the module stamps a single date, ageing is measured from data entry and every crane finding looks fresher than it is. Insist on separate found-date and entered-date fields, and report ageing from found-date. The gap between the two is itself a useful indicator of which sites have a working field-capture process.

The findings with no code owner

A large share of steel plant integrity findings sit outside any inspection code. Water-cooled stave or panel hot spots, refractory-backed shell temperature excursions, dust-handling erosion, molten-metal splash damage to structure, cooling water leaks into a furnace shell — none of these belong to ASME, API or AWS. They are governed by an internal engineering standard, if anything, and they are frequently among the highest-consequence items on the plant.

Because there is no code owner, these findings tend to live in shift logs and email. They never age, never aggregate, and never appear in the group integrity report, which then shows a plant whose only issues are pressure piping. A deficiency module that requires a code reference on every record actively excludes them. What it needs instead is a finding type where the governing basis can be an internal engineering standard or an operating limit, with the document reference and revision recorded like any other criterion.

This matters most for the standardisation job because these are the findings where site practice diverges hardest. One mill has a written stave temperature limit with a documented response; another has an experienced furnace operator. Bringing both into one register does not force the second site to write a standard overnight, but it does make visible that it has not, which is usually the first useful output the group gets from the exercise.

Deferral, fitness for service and the campaign calendar

Steel maintenance is campaign-driven in a way that continuous process plants are not. A blast furnace reline is a multi-year interval event. A caster segment change, a reheat furnace outage or a mill stand rebuild define the only realistic access windows for large parts of the asset base. Findings therefore get deferred far more often than in a refinery, and the deferral is legitimate — but only if it carries a technical basis.

The module has to make deferral a structured act rather than a status change. That means a required expiry date, a named approver with an authority level, and an attached basis: a remaining-life calculation from the measured wall loss and corrosion rate, a fitness-for-service assessment, or a documented engineering judgement. NBIC NB-23 Part 2 governs how the eventual repair or alteration is executed and by whom; it does not authorise the wait. The plant does, and the record of that decision is what an insurer, a regulator or an incident investigator will ask for.

Ageing then has to reflect this reality or the whole report is unfair. Report elapsed days since discovery, and separately report how many accessible outage windows a finding has survived. A finding eight months old with no access window since discovery is not a governance failure. A finding that has passed through two accessible outages is, regardless of its age in days, and that is the number the group review should be looking at.

Rolling out without a big-bang rename

The most common implementation mistake is to design the corporate standard first and migrate every site into it in one cutover. Steel groups rarely survive this, because the sites that resist are usually the ones with the most mature local practice and the most history to lose. The alternative that works is a translation layer: each site keeps its codes, its scale and its location names, and the module maintains a mapping to the corporate model that is visible, versioned and owned locally.

Sequence the rollout by data quality rather than by plant size. Start with the site that already has evidence-based closure and a clean functional-location register, because it will produce a working corporate report inside a quarter and give the programme a reference implementation. Bring the weakest site in last, after the mapping conventions have been proven, and expect its first pass to be a data clean-up exercise more than a software deployment.

Set the success test before you begin, and make it specific. A reasonable one: for any figure on the group dashboard, an integrity engineer can drill from the corporate number to the site record, see the original local severity, the mapping version that produced the corporate grade, the closure evidence, and who approved any deferral — in under two minutes and without contacting the plant. If the module cannot do that, the group has replaced three inconsistent registers with one inconsistent register.

How to evaluate a deficiency module for a steel group

Bring real records to the demonstration, not a scenario. Take one crane trunnion MT finding, one blast furnace gas main thickness set, one reheat furnace skid creep observation and one stave hot spot, each from a different plant and each in its native format. Ask the vendor to load them and produce a single ranked queue. The failures show up immediately: findings with no code owner get rejected, thickness sets lose their grid, and the crane record loses its found-date.

Then test the explain path in the other direction. Pick one number on the resulting dashboard and ask to be taken back to the constituent records with their original local grades and mappings intact. Ask what happens when Site B changes its severity scale next year — whether the history re-derives or freezes, and whether anyone can tell which mapping version produced a given historical report.

Finally, test the boring things that decide whether the system survives contact with a mill: does field capture work with no signal in a melt shop and sync later without duplicating records; can a technician attach an image and a technique sheet from a phone in gloves; and can a site administrator add a finding type without a vendor change request. Atlantis builds inspection management and reporting software on Odoo, configured to the plant's own taxonomy and severity model rather than a fixed one — affordable, accessible and fully customisable. Request a demonstration with your own multi-site records at info@atlantisndt.com.

Request a consultation