The Duplicate Weld IDs

During a post-construction documentation review for a privately held midstream company backed by private capital, we cross-referenced NDE examination reports against weld maps and pipe tally across several completed projects. Nearly a third of the weld identifiers in the record set appeared more than once.

Not a typo here and there — the same ID assigned to different physical welds, repeatedly, across multiple years of construction. The client's own investigation traced the source to the NDE contractor's reporting process. Every duplicated ID quietly split the record chain: two joints now shared one identity, and any document referencing that ID — an examination report, a welder log, a repair record — could belong to either one.

No single controlled register validated identifier uniqueness — the one integrity check that every downstream record depends on.

The weld ID is the primary key of pipeline traceability. Everything hangs from it: the weld map, the welder's qualification continuity, the NDE examination, the pressure test documentation, the as-built location. Duplicate the key and every record attached to it becomes ambiguous — and in a records context, ambiguous means unverifiable.

This is not a paperwork quibble. After the 2010 San Bruno rupture, the NTSB's urgent recommendations and subsequent PHMSA rulemaking established that operator records must be “traceable, verifiable, and complete” — now codified at 49 CFR §192.12. A duplicate ID fails all three tests at once: it isn't traceable (which joint is this?), it isn't verifiable (which examination is real?), and the record set isn't complete (the chain is broken). Under §192.619, segments lacking such records must be reconfirmed with conservative assumptions — records failures convert directly into operational and financial consequences.

If you cannot prove which weld an examination report documents, then as far as a regulator, an integrity engineer, or a court is concerned, you may have no record of that weld's examination at all.

The root is records that exist but cannot be trusted — present, filed, and wrong in fact. The public record shows what that root produces:

Different decades, different states. The constant: records that were present, filed, and could not be trusted when it mattered.

The cost of this defect did not stay with the operator, and the fight over it became a cost of its own. The NDE contractor reissued reports and defended its reporting process, likely absorbing re-examination expenses under back-charge. The inspection firm's oversight work came under question — and prequalification files feel that for years. The operator paid for reconciliation regardless of fault, because the asset is theirs. And on the next project, every bidder inherited tighter data submittal requirements: a trust tax the entire chain pays for one firm's process failure.

Here's the uncomfortable part: these projects had welding inspector oversight throughout. But oversight of welding is not validation of records. Inspectors inspected — but nothing in the process checked identifier uniqueness, because no single controlled register existed to check against. Each project generated documents; nobody reconciled them against each other.

The duplication survived three levels of participants: the contractor's reporting process, the inspector oversight layer, and the client's records acceptance. Multiple years of records accumulated the defect because every level assumed another one was checking.

Remediation required reconciling every duplicated identifier against weld maps, tally sheets, and examination reports, joint by joint — archaeology on records that were supposed to be usable as-is. Where provenance can't be reconstructed, the fallback is worse: re-examination, re-review of test documentation, and in the worst case, treating the weld as undocumented.

None of this would have been necessary if the duplication had been caught the day it happened, when the fix is a single rejected entry. Retroactive review always costs more than concurrent capture — this case is a clean example of why the gap between construction and review is where record defects compound.

Uniqueness of identifiers is the cheapest integrity check there is — and one of the most expensive to recover from when it's been missing for years. If no one in your process is validating it, it isn't happening.