TURNAROUNDQC
← Back to Turnaround QC
Last updated 12 August 2026
A turnover package is the argument that a system is fit to go back in service, made to somebody who was not there and is not going to open anything up to check. This is what goes in one on a turnaround, section by section — worked through a single crude unit rather than described in the abstract.
The job spec governs. Section list, format, signatories and retention come from your contract and the governing code. What follows is the shape most turnaround packages take when nothing tighter has been specified, and the order is the order an owner’s reviewer actually opens things in.
Crude unit turnaround, twenty-one days. The C-1101 atmospheric tower opened for internal inspection with two tray decks renewed. The E-1101 A/D preheat train with all four bundles pulled, cleaned and inspected, and E-1101C retubed after the inspection found what it found. Fourteen field tie-ins across the overhead and kerosene side-cut lines. Every relief valve on the unit pulled, bench tested and reinstated. Everything hydrotested and handed back system by system.
Two hundred and some records. Small enough to follow end to end, large enough that every failure below has somewhere to hide.
This is the thing that makes a turnaround package different from every other kind, and it is the thing most often got wrong by people arriving from construction or from maintenance.
A maintenance package is assembled per job. A construction package is assembled per contract, per area, per subcontractor. Neither is any use to the man who has to decide whether he can put hydrocarbon back into the overhead line on Tuesday night. He is not accepting your contract. He is accepting a system — a defined set of lines, vessels, valves and relief devices with a boundary drawn on a marked-up P&ID — and he needs every record that bears on that boundary in one place, whoever did the work and whatever purchase order it sat under.
So the package is built per system from the first day, not sorted into systems at the end. On this job:
| System | Boundary | What is in it |
|---|---|---|
| CU-01 | Tower and overhead, tower skirt to the overhead condensers | C-1101 entry, internal inspection, two tray decks, eight tie-ins, four PSVs |
| CU-02 | Preheat train, crude pumps to the heater inlet | E-1101 A/D bundles, the C retube, shell and channel joints, two PSVs |
| CU-03 | Kerosene side-cut, drawoff nozzle to rundown battery limit | Six tie-ins, one section of thinned line renewed, three PSVs |
| CU-04 | Unit relief and flare header | The remaining PSVs, tail pipes, header restoration |
The second consequence is the one people argue about. Mechanical completion is granted per system with its punch list attached, not once the punch list is empty. Waiting for an empty list means nothing is ever handed back, because on the last week of a turnaround there is always a lagging item, a missing nameplate, a temporary support. The owner accepts CU-02 with eleven open punch items because the certificate says which eleven, which category each is in, and who owns each one. A clean certificate with an unstated backlog behind it is worth less than a dirty one that tells the truth.
The cover names the unit, the system, the turnaround, the package revision and the date. A package with no revision on it cannot be superseded, only argued about — and on a turnaround it will be superseded, because packages go in before the last punch item closes.
The index is the load-bearing part. It lists every section, the number of records in each, and an explicit not applicable where a section does not apply to this system. An absent section and a section with nothing in it look identical from outside, and the reviewer has to assume the worse of the two.
What an experienced reviewer opens first is not the cover. It is the isolation reconciliation, then the mechanical completion certificate with its punch list. Those two answer the only question he has been asked, which is whether the unit is safe to fill. Everything else is him checking whether he believes the answer.
A marked-up P&ID with the boundary drawn on it, the line list for the system, and the work list. Without the marked-up drawing, nobody can tell whether a record belongs in this package or the next one, and tie-in welds on a system boundary end up in both packages or in neither.
The work list is split. Planned is what was on the approved list at scope freeze. Discovery is what was found once equipment was opened — the E-1101C retube is discovery, because the bundle was pulled to be inspected and the inspection condemned it. Emergent is work the execution created: a weld cut out, a failed test, damage in the shop. Owner-added is what came in after freeze.
Held as one bucket the growth tells you only that it is growing. Split, it says something different in each case, and the categorisation has to be made when the job is raised. Nobody re-categorises three hundred work orders honestly at the end of a turnaround, and the only person who reliably knows why a job exists is the person who raised it.
The register is read early because it can end the review. It is one date comparison per instrument and it invalidates everything downstream of a failure.
MTRs for every pressure-retaining item installed, matched to the heat numbers actually in the ground. The chain has to survive the saw: a heat number stencilled on a length of pipe is worthless once that length is six pieces, so marking is transferred before it is cut and the transfer is witnessed and recorded. On the kerosene renewal that is one twenty-foot length becoming four spools and eight welds, and the heat number has to be findable against every one of them.
Where an alloy is specified, PMI, to the extent the spec requires — and the weld deposit as well as the base metal, because wrong filler between two correct components is the version nobody finds. Every NDE method there is will pass a sound weld made in entirely the wrong material. None of them asks what the metal is.
The heart of the piping side, and on this job it covers fourteen tie-ins plus the kerosene renewal.
Two of the overhead tie-ins are closure welds that cannot be hydrotested, because pressurising against them would mean pressurising equipment that is already reinstated. Where the code and the owner permit examination in place of a test, the examination becomes the only evidence that joint will ever have. Those welds get named as such in the package, with the approval that allowed it. They should not be quietly indistinguishable from the twelve that were tested.
Every joint broken for the work and every joint made up: gasket type, size and rating; stud material and whether new or reused; lubricant; the target and where the target came from; method — hand torque, hydraulic, tensioner — the pass sequence, and the identity of the calibrated tool.
Torque is not clamp load. Most of the torque goes into friction and the friction is set by what is on the threads, so a figure with no lubricant beside it describes a wrench setting rather than a joint. And the joint that leaks on start-up is usually not the badly torqued one. It is the one nobody recorded breaking — opened by scaffolders or the blind crew for access, never entered on the register, never on the make-up list, never tested.
Break and make-up are two events with two dates and two names. On a turnaround the crew that opened the E-1101C channel on day four is not the crew that closed it on day seventeen.
The test package is assembled before anything is pressurised: the test procedure, a marked-up P&ID showing test limits and blinds, the calculated test pressure with the design conditions it came from, the vent and drain arrangement, and the list of items excluded or removed — relief valves, control valves, instruments, anything that must not see test pressure.
The record itself carries the medium and its chloride content where the material demands it, ambient and metal temperature, two calibrated gauges of different ranges with both serial numbers and both due dates written on the sheet, the hold period, the pressure at start and end of hold, and the signatures of the parties who witnessed it. A gauge that has quietly drifted reads exactly like one that has not, which is the whole reason for the second gauge, and a certificate in a folder proves a gauge was calibrated but not that that gauge took this reading.
Then depressurisation, draining, flushing and dewatering, recorded with the same seriousness as the test. The failure the drying record prevents does not happen on test day. It happens in a dead leg over the following summer.
Every blind fitted for this system, by tag number, with fitting and removal as two separately signed events — date, time and name on each, not one field that flips from in to out. What was fitted and what replaced it: a spade pulled with the spacer not refitted leaves a joint that closes short.
The register is reconciled against the physical tag board. Tags out counted against tags in, walked, with spectacle blind orientation read from the ground wherever it can be. The asymmetry is structural rather than careless: fitting a blind is a permitted job that blocks the work behind it, so the record makes itself. Pulling it is one line on a checklist at four in the morning with nothing waiting on it.
This is the section the reviewer opens first, and he is right to. A missing blind removal signature is the one defect in the package that can hurt somebody on start-up day, and finding it while the flange is still reachable costs an hour rather than the start-up.
Reinstatement covers what went back after the test: relief valves refitted with the correct tail pipe and car seal position, control valves and orifice plates reinstalled the right way round, instruments reconnected, spades pulled, insulation and painting restored, temporary supports removed and permanent supports and spring hangers released and set to their cold positions.
Then the walkdown, by the owner and the contractor together, against the marked-up P&ID. Items are raised and categorised. Categories are project-defined, but conventionally: A blocks mechanical completion or start-up, B must be closed within an agreed period after start-up, C is cosmetic and closes at leisure. What matters is that the categories are defined in writing at the start of the job and applied by the same rule on the last night as on the first walkdown.
The mechanical completion certificate names the system, states the standard it was completed to, lists the documents it relies on, and carries its open punch counts on its face: zero A, eleven B, six C. It is signed by the contractor, the owner’s representative and, where the work touched pressure-retaining items in service, the inspector. That certificate is the deliverable. Everything upstream of it exists so that somebody can sign it and mean it.
None of these require any technical knowledge to run, and none require anybody to look at the plant. That is exactly why they are the first things a reviewer does.
| The comparison | What it means |
|---|---|
| Instrument due date against the date it produced a reading | Every number that instrument produced is void back to its last known good calibration — including the ones that passed. One sticker can take out a fortnight of tests across several systems. |
| A weld repair with no re-examination | The defect is recorded, the fix is recorded, and there is no evidence the fix worked. The package has documented a known defect and stopped. |
| A hold point released with no notification date | A hold point stops the work until the nominated party attends or releases it. Without an attendance, a written release or a logged notice, the honest case and the skipped case produce the same blank. |
| A blind with an install signature and no removal signature | Either the blind is still in the line or the paperwork is wrong, and from outside those are identical. The reviewer must assume the first. |
| An NDE result with no named acceptance criterion | Not a result. Acceptance differs by code, edition and fluid service category, and “acceptable” against nothing cannot be checked by anyone. |
| A PSV reinstalled with no as-found pop | The one piece of evidence about how that valve behaved in service has been discarded, and it cannot be recovered by testing it again. |
Two more worth running as a pass of their own: dates that run forward through the package — entry before internals, internals before box-up, box-up before test, test before reinstatement — and the tag spelled identically everywhere it appears. E-1101C here and E1101-C there means nothing can be pulled together at the end.
It is read by one person, under time pressure, deciding whether to accept a unit back and put hydrocarbon in it. He was not there when the work was done. He is not going to break a flange to check. Everything in the package exists so that he can test a claim without taking anything apart: the claim that the weld is sound, that the metal is the right metal, that the bolts are tight, that the tray is level, that the relief valve will lift, that the blind came out.
Which is the standard to hold a record to while you are making it, at four in the morning on day nineteen with the crane booked. Not whether it is filled in. Whether somebody who was not there could use it to check the claim it makes.
That is what this app compiles, system by system.
What goes in a turnaround turnover package — turnaroundqc.com/turnover-package