Getting One Set of Thickness Numbers Out of Several LNG Sites That Each Count Differently

The rollup fails before the software is even involved

Corporate integrity leadership at a multi-site LNG operator asks a simple question roughly quarterly. What percentage of our condition monitoring locations are current, and where are we exposed. The question is answerable at each site individually. It becomes unanswerable the moment the three site answers are added together, and the reason is not that the sites use different databases.

It is that they use different definitions. Site A counts a location as an area on a fitting containing four clock-position readings. Site B counts each clock position separately. Site C, which inherited its register from a construction contractor, registered every weld in a circuit as a location and then never read three quarters of them. Adding these produces a number, and the number is meaningless.

Every subsequent fix is downstream of this. Deploying a common system across sites that keep their own definitions simply centralises the incoherence. The work is defining the object, the location, the measurement point, the circuit, the mechanism, the minimum thickness basis, and then migrating each site onto those definitions with its legacy identifiers preserved as aliases so nobody loses their history.

Cryogenic service breaks the assumption the register was built on

Thickness monitoring exists because carbon steel in a corrosive process loses wall at a rate that can be trended and extrapolated. That model holds well across a refinery. It holds badly across an LNG plant, because a large fraction of the plant runs dry, sweet hydrocarbon at temperatures where general corrosion is negligible and the metallurgy is austenitic stainless, aluminium or 9% nickel steel.

The credible threats in those sections are different in kind, not just in rate. External chloride stress corrosion cracking under cold insulation on austenitic stainless, particularly where lines cycle through the moisture condensation range during trips and warm-ups. Mechanical damage and thermal fatigue at cold-warm transitions. Liquid metal embrittlement of aluminium if mercury breaks through the removal bed. None of these announce themselves as a thinning trend, and none are caught by an ultrasonic spot reading.

A registry that only knows how to hold thickness therefore forces a bad choice on the site. Either register cryogenic equipment with thickness locations that will read nominal forever and create false assurance, or leave it out and create a blind spot. The correct design registers the location, records the applicable damage mechanism using API 571 terminology, and attaches the inspection method appropriate to that mechanism, which may be insulation removal and visual, dye penetrant on stainless welds, or a documented surveillance regime under API 583.

Where LNG wall loss really is, and why it clusters on the warm end

The thickness program in an LNG facility earns its keep on the pretreatment train and the utilities. Amine treating for acid gas removal carries the classic set of mechanisms described in API 571: amine corrosion concentrated where rich amine flashes, at the regenerator overhead, and across let-down valves, along with amine stress corrosion cracking in non-post-weld-heat-treated carbon steel. This is the area where a poorly placed grid genuinely misses metal loss.

Dehydration and molecular sieve regeneration add a thermal cycling dimension. Damage accumulates with regeneration cycles rather than with elapsed calendar time, which means a unit that has been swung hard during a demand year has aged faster than the calendar suggests. A registry that only schedules on dates cannot express that, and the site engineer compensates informally, which is precisely the sort of undocumented local practice that breaks a fleet rollup.

Utilities are the quiet one. Seawater cooling, fire water and treated water systems corrode steadily and generate a large share of the total location count. If they sit in the same pool as process locations, a site can raise its headline coverage number simply by reading easy utility points, and the fleet view rewards the wrong behaviour. Ring-fencing them is not cosmetic.

Minimum thickness is where standardisation gets political

Two sites can hold identical readings on identical spools and disagree about remaining life by a decade, because they disagree about what the reading is being compared to. One computes t-min from the pressure design formula in ASME B31.3. Another applies a structural minimum for large-bore, low-pressure piping where the pressure calculation returns an implausibly thin wall. A third carries an owner-specified retirement thickness set above both by a previous integrity manager as a conservatism.

All three are defensible positions. What is not defensible is a fleet remaining life distribution built by pooling them, because the distribution then encodes three different risk appetites and cannot be used to allocate capital. This is the point in a standardisation project where the discussion stops being about software and becomes a corporate engineering decision that someone has to own and sign.

The registry's contribution is to make the divergence impossible to hide. Every location should carry its t-min basis as a governed field, and any rollup that spans more than one basis should either segment by basis or refuse. A system that silently pools them is not neutral. It is generating a number that a senior engineer will later be asked to defend and will not be able to.

Governing the standard without erasing what each site already has

The fastest way to kill a standardisation programme is to renumber everyone's locations. Site engineers know their plant by the identifiers painted on it and printed on twenty years of reports. A migration that replaces them produces a period during which nobody trusts either the old or the new labels, and the field crews quietly keep using the old ones on their sheets.

The workable pattern is a governed corporate identity with unlimited site aliases. The corporate identifier is what the fleet reports on. The site's legacy identifier, the construction contractor's tag and the isometric callout all persist as searchable aliases resolving to the same record. A technician typing the number stencilled on the pipe finds the location. A corporate analyst querying the fleet gets one identity.

Standards also change, so the standard itself must be versioned. When the definition of a condition monitoring location is revised, historical records must remain interpretable under the definition in force when they were created, with an explicit mapping forward. Without that, the first revision destroys the ability to compare this year to last year, and the programme loses its main proof of value.

What a comparable coverage metric actually looks like

A defensible fleet metric names its unit and its scope in the metric itself. Percentage of process condition monitoring locations, counted at location level, with a due date derived from the corporate interval rules, currently within interval. That sentence is longer than the number, and it should be, because every clause in it is a place where two sites can diverge.

Second, the metric should be reported alongside its denominator, not just as a percentage. A site improving from eighty to ninety percent by retiring locations rather than reading them is a familiar pattern, and it is only visible when the count is shown. Retirements should require a reason and an approver, and should appear in the report as a separate line rather than vanishing into the base.

Third, the metric needs an explicit not-applicable category for equipment under a non-thickness surveillance regime. Cryogenic piping under cold insulation surveillance, the inner tank of a full containment storage tank under API 620 Annex Q, and mercury-exposed aluminium are all monitored and none of them belong in a thickness percentage. Reporting them as a named category is the difference between a metric an engineer trusts and one they work around.

The regulatory audience is not the same as the refinery one

United States onshore LNG facilities sit under 49 CFR Part 193, which incorporates NFPA 59A, and the operating and maintenance obligations there are framed differently from the mechanical integrity language a refinery integrity engineer is used to under OSHA 29 CFR 1910.119. Facilities that export or serve marine terminals also carry Coast Guard obligations, and internationally EN 1473 shapes the equivalent expectations.

Practically, this means the evidence an LNG operator has to produce is more about demonstrating a controlled, documented and followed procedure than about defending a specific corrosion rate calculation. That plays to the registry's strengths: an auditable record of who defined a location, on what basis, when it was read, by whom, with what equipment and calibration, and what was done about the result.

It also means the fleet standard has to accommodate sites in different jurisdictions without forking. A regas terminal in one country and a liquefaction train in another will have different statutory inspection triggers layered on top of the same engineering basis. The registry should treat the statutory trigger as an additional constraint on the due date rather than as a separate parallel schedule, so that a location has one due date and a stated reason for it.

How to evaluate a registry for a multi-site LNG estate

Give the vendor two real extracts from two of your sites with their genuine, incompatible conventions, and ask them to produce one comparable coverage number without editing the source data by hand. What you are watching for is whether the system exposes the incompatibility and asks for a governance decision, or silently produces a number. The second behaviour is disqualifying, and it is common.

Ask to see the alias resolution work in the field capture interface. Type the legacy site identifier, the construction tag and the corporate identifier, and confirm all three land on the same location record with the same history. If the field tool only recognises the new corporate number, the crews will not adopt it and the register will drift within one campaign.

Finally, ask the system to represent an asset that legitimately has no thickness programme, such as a full containment inner tank, and show how it appears in the fleet report. A system that can only classify it as overdue, or can only handle it by leaving it out, will force one of your sites to lie in order to look correct. Atlantis will run this exercise against a sample of your own multi-site data during a scoping consultation arranged through info@atlantisndt.com.

Request a consultation