Projecting corrosion to the end of the next run, before the scope freeze

The turnaround changes the question the corrosion rate has to answer

Outside a turnaround, a corrosion rate answers how long a component has left and when to look at it again. API 510 and API 570 turn that into an interval, the lesser of half the remaining life or a class maximum. It is a rolling question, it is revisited every time new data arrives, and being a few months late to the answer costs very little in practice.

Inside turnaround planning the question changes shape entirely. A petrochemical unit's window is fixed years in advance and is frequently immovable: an olefins complex shuts on a date driven by contractual obligations to downstream polymer plants, and the cracker's run length is itself a commercial commitment rather than an engineering preference. The question becomes binary and dated. Does this component clear the next run, and have I decided in time to buy the metal.

That reframing is what most corrosion rate tools miss. They report remaining life in years measured from today. A turnaround planner needs projected thickness at a specific future date, computed on the governing rate, with the confidence in that rate exposed rather than hidden, because the cost of being wrong is asymmetric in both directions and it lands on a critical path rather than on an inspection interval.

Projecting to the end of the next run, not to today

The projection itself is simple arithmetic. The exposure period is the part people get wrong. It is not the time to the turnaround; it is the time to the turnaround plus the entire following run, because whatever is not replaced during the window has to survive until the next one opens. A unit eighteen months from a shutdown with a sixty-month run length is asking its metal for six and a half more years of service, not eighteen months.

With the period fixed, both rates are projected and compared against the retirement thickness. Where the long-term and short-term projections land on opposite sides of that line you have a genuine scope decision that turns on which rate you believe, and that is exactly the population deserving engineering attention and usually a confirmatory scan before the freeze. It is normally a small fraction of the CML population, which is what makes it tractable.

Where both projections land above the retirement thickness there is no thickness scope and the resource should go elsewhere. Where both land below, the decision is already made and the only remaining question is procurement lead time. Sorting the whole CML population into those three buckets automatically, against a dated projection rather than a rolling remaining life, is the single most useful thing this module does for a turnaround team.

Calendar time is not exposure time in a petrochemical unit

A corrosion rate expressed per calendar year carries an assumption that the metal was in service, at design conditions, for the whole of that year. In refining that assumption is usually close enough to true. In petrochemicals it frequently is not. Crackers cut rates. Trains are idled through a demand trough. A polymer line runs a different grade for a quarter. A unit sits on hot standby through an upstream outage that had nothing to do with it.

The arithmetic consequence runs in one direction and it is the unsafe one. If a circuit was in aggressive service for eight of the last twelve months, the true rate per operating year is roughly fifty percent higher than the calendar rate, and the calendar rate is the number that goes into the projection. The next run, if it is a full one, then consumes wall faster than the projection allowed for, and the error is discovered in service rather than in a review.

The remedy is to hold the operating history alongside the readings: campaign, feedstock, operating window and periods out of service. The rate can then be expressed per operating year and projected against the run that is actually planned, and a circuit whose damage is campaign-specific can be projected against the campaigns scheduled rather than against an average of the past. This is also the data that makes an API RP 584 integrity operating window meaningful, because a window that was never approached during an idle period was never actually tested.

Petrochemical damage that a thickness rate will never show

Much of what shuts a petrochemical unit down is not general wall loss. Caustic stress corrosion cracking in wash and scrubbing services, chloride stress corrosion cracking in austenitic stainless where insulation has held chlorides against a warm wall, carburisation and creep in cracking furnace tubes, and polythionic acid cracking that develops during a shutdown are all mechanisms where the thickness trend reports a comfortable zero right up until the component fails.

This matters for turnaround scoping specifically, because thickness data is abundant and cracking data is not. A scope built from the corrosion rate module alone will be dominated by the mechanisms that happen to be measurable by a thickness gauge, and it will systematically under-scope the ones requiring a different technique: wet fluorescent magnetic particle examination, phased array, eddy current on tube bundles, or in-situ metallography and replication.

The practical arrangement is to carry the credible damage mechanisms on the circuit record next to the rate, per API RP 571, and let the scope screen work off both. A circuit with a 2 mpy rate and a known chloride cracking susceptibility should surface in the scope for reasons that have nothing to do with its remaining life, and the module should be capable of saying so rather than filing it silently under no action required.

The scope freeze is the real deadline, not the inspection

Every turnaround runs through a sequence of gates: scope development, scope freeze, long-lead procurement release, contractor award, detailed planning, then the window. For a major petrochemical shutdown the freeze is typically nine to twelve months out, and long-lead material release can be earlier still where alloy pipe, clad heads or exchanger bundles carry lead times measured in tens of weeks.

The corrosion rate therefore has a deadline attached to it, and that deadline is frequently earlier than the next planned inspection. A CML whose next survey falls two months after the freeze is, for scoping purposes, blind: the decision will be taken on data that is already a run old. Knowing which CMLs sit in that position, early enough to do something about it, is what allows a handful of on-stream surveys to be pulled forward while the result can still change an outcome.

So the module needs the turnaround calendar inside it rather than beside it. Each projection carries the decision date it feeds, each circuit shows whether the data supporting its decision will be refreshed before or after that date, and the ones that will not be refreshed are the ones the integrity engineer works first. That is a scheduling function as much as a calculation, and it is the reason the rate engine belongs in the same system as the inspection plan rather than in a spreadsheet next to it.

Long-lead materials, access and crew: what the governing rate actually buys

A scope line is not a line. It is a bill of materials, a scaffold, an insulation removal footprint, a confined space or hot work permit, a crew, a position on the critical path and often a repair or re-rating calculation that has to be engineered before the window opens. Deciding it on the short-term rate rather than the long-term rate can commit a procurement package many months before anybody will be able to confirm whether it was needed.

Deciding it the other way is worse. A component discovered thin during the window becomes an emergency: expedited material at a premium, a work front inserted into a schedule with no slack, and in the bad case a unit that returns late while downstream polymer plants run short of feed. That asymmetry is why the conservative rate governs by default in a turnaround context even more firmly than in a routine interval calculation.

What the module can do is shrink the population where the asymmetry bites. Every CML it moves out of the disagreement bucket and into agreement is a procurement decision taken off the table. That is achieved through data quality and through targeted confirmatory inspection ahead of the freeze, and a scan on a dozen contested circuits is very cheap set against the alloy orders it either releases or cancels.

Evaluating the module against your own turnaround calendar

Load a real unit and a real shutdown date. Ask the system to produce, for that date and your planned run length, the three buckets: clears on both rates, fails on both, and disagrees. If it cannot do that without an export into a spreadsheet, what you are looking at is a records system rather than a planning tool, and the spreadsheet will end up being the thing your turnaround actually runs on.

Then examine the second-order behaviour, because that is where these tools separate. Does it flag CMLs whose next survey falls after the freeze? Does it distinguish calendar years from operating years when the unit had an outage? Does it carry credible damage mechanisms so a cracking-susceptible circuit is not filtered out by its low rate? Does it record who accepted a governing rate, and can it reproduce a projection made six months ago using the inputs that existed then?

Finally, run it backwards on your last turnaround. Take the data as it stood at that scope freeze, produce the projection, and compare it against what you actually found once the equipment was open. That single exercise reveals more about how much confidence your rates deserve than any vendor demonstration, and it is a test you can run against your own history before you commit to anything. Contact info@atlantisndt.com to arrange a demonstration using a unit of your own.

Request a consultation