What a failed pressure test taught us about PWHT thermocouple placement

Ananya WatanabeMetallurgical Failure Analysis · 25 yrs

This is the most common PWHT failure mode we see in third-party review and it almost never gets caught by the paperwork, because the chart is genuinely compliant — the thermocouple did reach soak. The specification gap is that most PWHT procedures specify the soak temperature and hold time but say nothing about where the controlling thermocouple attaches relative to section thickness. Your fix is exactly the right one. The detail I would add: specify not just the thickest section but a minimum number of thermocouples with at least one on the thickest section and one on the thinnest, because a procedure that only protects the thick section can overshoot the thin one and you trade a hardness spike for a softened, over-tempered thin section.

Worth noting how far the heat flow explanation goes. A 44 mm section against an 18 mm section is not quite a factor of six in thermal mass per unit area, but the through-thickness lag on local electrical resistance heating is dominated by the heated surface to volume ratio and the effect is large. Adding a soak time margin without moving the thermocouple does not reliably fix it either, because the steady-state gradient across the thick section can stabilise below soak if the insulation coverage is thin at that point. The placement is the control, and the hold time only helps once the thermocouple represents the governing thickness.

Jonas YilmazGas Processing Engineering · 24 yrs

Verified expertise in Gas Processing Engineering

The part of this that deserves more attention than the metallurgy is the investigation sequence, because it is a textbook case of testing the accessible hypothesis first. Welder technique and consumable are both cheap to investigate and cheap to change, so they got investigated and changed. Neither was ever supported by evidence from the failed part — the hardness traverse that actually identified the cause was available after the first failure and nobody ran it. We now have a standing rule that no corrective action is approved on a repeat failure until someone has sectioned the failed component itself, not a coupon.

The specific tell you had, which is worth writing down for others, is that a HAZ hardness spike above the WPS limit is nearly diagnostic. It says the thermal cycle that part of the joint experienced was not the cycle the procedure intended, and there are only a few ways that happens: the PWHT did not reach that location, the preheat or interpass was not maintained, or the material is not what the certificate says. Those three are all cheap to check and they narrow the problem immediately. Requalifying welders cannot produce that hardness spike, which means the first investigation was disproved by evidence that already existed.

Rohan TanPlastics Engineering · 20 yrs

Adding a practical note on witnessing, since your fix relies on it. Witnessing the first three production welds after a procedure change works, but only if the witness is looking at the right thing. Ours were initially witnessing the welding and signing the traveller, which catches technique problems and misses thermocouple placement entirely because the placement happens hours later with a different crew. We moved the witness point to thermocouple attachment and insulation coverage, before power on, and included a photograph of the attachment against the thickness transition in the record.

The photograph turned out to be the highest-value part of the whole change. It is a thirty-second step, it is unambiguous, and it means a reviewer six months later can confirm placement without taking anyone's word for it. We have since found two placement errors from the photographs alone, both on jobs where the chart was compliant and would have passed review. Neither reached service.