Case Study: Crude distillation tower, 18 ft ID, SA-516 Gr 70, 38 mm wall — Petroleum refining (crude distillation)

Asset & context

Petroleum refining (crude distillation) — US West Texas (Permian Basin). Crude distillation tower, 18 ft ID, SA-516 Gr 70, 38 mm wall. Crew of 4 for 8 days.

The challenge

The crude tower at a 165,000 bpd Permian Basin refinery was scheduled for its 10-year ASME VIII Div 2 internal/external integrity inspection during the spring 2026 turnaround. The tower lower-shell-to-bottom-head girth weld had been repair-welded in 1991 after a fab-shop NDT rejection — the original weld had failed acceptance for cluster porosity. The 1991 repair was 100% RT inspected per ASME VIII Div 2 Mandatory Annex 7-200, signed off, and the tower entered service.[1] The repair had been re-inspected by film RT during the 2006 and 2016 turnarounds, both times passed. For the 2026 turnaround, the asset integrity team added baseline TOFD on all original girth welds and repair welds as a pre-condition for extending the inspection interval to 12 years. The TOFD scan on the 1991 repair weld returned a 38 mm planar reflector signature 22 mm below the OD surface — the size and location pattern of lack-of-side-fusion, a defect class that film RT struggles to detect when the defect plane is parallel to the radiation beam.

Approach

Method selection: TOFD as the planar-defect baseline complement to RT

RT and TOFD are complementary, not interchangeable, on heavy-wall pressure vessel girth welds. Film RT per ASME V Article 2 has excellent probability of detection (POD) for volumetric defects — porosity, slag, gas pockets — but POD drops sharply when the defect plane is parallel to the radiation beam.[2] Lack-of-side-fusion (LOSF) in a thick-wall multi-pass weld is the classic problem case: the defect is a planar reflector aligned with the fusion line, often only 0.1-0.3 mm gap, oriented at roughly the bevel angle relative to the OD surface. Conventional RT can miss LOSF entirely on a 38 mm wall regardless of film quality or exposure parameters.

TOFD per ASME V Article 4 Mandatory Appendix III is specifically calibrated for this defect class.[3] TOFD uses paired transmit/receive probes at a fixed offset, capturing diffracted signals from the tips of planar defects. A 38 mm LOSF at the bevel angle gives a strong tip-diffraction signature that is unmistakable in the TOFD A-scan and B-scan. POD for planar defects ≥ 5 mm length in 25-50 mm wall is typically > 95% with a properly qualified TOFD procedure per ASME V Article 4 §T-471.[3]

The decision logic in the inspection plan was: RT remains the primary acceptance method for new construction and field welds per ASME VIII Div 2 7-200. TOFD is added as a baseline scan during the first scheduled outage after a repair, and is mandated for any weld with a known repair history. PAUT was specified as the secondary method if TOFD found a reportable indication — PAUT sizing per ASME V Article 4 Appendix IV would give the depth and through-wall extent needed for fitness-for-service evaluation.[4]

Procedure: TOFD scan plan, calibration, and indication characterization

The TOFD scan was performed externally on the 38 mm-wall lower-shell girth weld with a 5 MHz, 12 mm element TOFD probe pair at 60° refracted angle, separation of 50 mm probe centers. The probe pair was mounted on an encoded girth-band scanner around the full 18 ft tower circumference. Calibration on a 38 mm-thick mockup block with side-drilled holes per ASME V T-472 confirmed sensitivity to 80% screen response on a 1.5 mm SDH at mid-wall depth.[3] Lateral wave clarity was verified at the start of each scan quadrant — lateral wave amplitude variation across the scan stayed inside the ±2 dB acceptance band.

Scan speed was 80 mm/sec at 0.5 mm encoder resolution, generating a continuous B-scan around the full 17.2 m circumference in 4 minutes of scan time plus 90 minutes of setup, calibration, and surface prep. The TOFD B-scan was analyzed in real-time by a Level III TOFD inspector against ASME V §T-475 acceptance criteria.[3] An indication 38 mm in length, 22 mm below the OD surface, oriented along the fusion-line geometry of the original bevel, was identified at clock-position 4:20.

PAUT follow-up was performed at the indication location with a 5 MHz, 64-element linear array at 0.6 mm pitch, scanned in S-scan mode from 35° to 70° refracted angle. The S-scan resolved the defect as a planar reflector 38 mm long, 11 mm through-wall extent (from 22 mm depth to 33 mm depth), aligned with the original weld bevel face. The defect signature was consistent with lack-of-side-fusion on the upstream side of the 1991 repair weld — likely caused by a low-heat-input root pass during the repair, with the side-wall fusion never re-established before the cap layers were laid down.

Findings: 35-year-old defect, no growth, no in-service damage

Two questions had to be answered before disposition: (1) Had the defect grown in service? (2) Was there any associated in-service damage (fatigue cracking from the LOSF tip, environmental cracking, HAZ degradation)? The PAUT sizing was compared against archived film-RT images from 2006 and 2016. While film RT had not "called" the LOSF, the 2016 film showed a faint linear shadow at the same clock-position when re-examined with the TOFD-confirmed coordinates in hand — consistent with the LOSF being a fabrication-era defect rather than a service-induced crack.

A confirmatory MT inspection of the OD surface above the defect per ASME V Article 7 found no surface-breaking cracks.[5] A wet-fluorescent MT scan of the ID surface during the internal walk-down found no surface indications. A TOFD re-scan at lower frequency (2.25 MHz) confirmed no through-wall fatigue crack growth from the LOSF tip — the defect signature was clean and undisturbed.

Fitness-for-service evaluation under API 579-1 Part 9 (Assessment of Crack-Like Flaws) was performed.[6] The 38 mm × 11 mm planar defect at 22 mm OD depth, in SA-516 Gr 70 base metal with the tower operating at 2.1 MPa MAWP and 360°C max process temperature, returned a critical flaw size of 95 mm × 18 mm at end-of-life conditions. The existing defect was well inside the critical-flaw envelope — but API 579 §9.4.3.2 requires demonstration that the defect will not grow to the critical size before the next inspection. Crack-growth analysis using a Paris-law approach for fabrication defects in pressure vessel service returned zero predicted growth, consistent with 35 years of in-service stability.

Disposition: weld repair vs FFS run-as-is — the cost-driven decision

API 579 §9.4 allowed the operator to run the tower as-is with the LOSF in place, on the engineering data. But the integrity team weighed three factors against pure FFS: (1) the 12-year inspection interval extension required a clean baseline, and the LOSF would complicate every future TOFD inspection at that location; (2) regulatory and insurance posture preferred a defect-free pressure boundary on a 35-year-old tower; (3) the tower was already opened for the turnaround, and a 38 mm × 11 mm weld repair was a 36-hour scope inside the existing critical path.

The repair was specified per ASME VIII Div 2 Part 6 §6.6.1 as an excavation, re-weld, and 100% volumetric NDT (RT + TOFD + PAUT) of the repair zone.[1] The defect was excavated by air-arc gouging to 2 mm past the deepest extent confirmed by PAUT. The cavity was MT-inspected before re-welding. SMAW with E7018 electrodes, preheat 200°C, PWHT at 620°C for 1 hour. Post-repair NDT: RT to ASME V Article 2, TOFD to Article 4 App III, PAUT to Article 4 App IV — all three passed.

The 12-year inspection interval extension was approved. Total inspection-and-repair cost: $238,000. The avoided cost scenario — if the LOSF had been left in place under FFS and had subsequently grown due to a thermal-cycle event not captured in the original analysis — sits in the range of a major incident, with the worst-case crude-tower rupture historically running $50-150M in damage, lost production, and regulatory action per industry incident data.

Methods deployed

Defects found

Outcome

Defect characterized as fabrication-era (1991 repair) with zero in-service growth over 35 years. API 579-1 Part 9 FFS confirmed sub-critical, but excavation and re-weld were selected to support 12-year interval extension. Post-repair RT + TOFD + PAUT all passed. Tower returned to service at MAWP.

Cost avoidance

Estimated $50M+ against a worst-case crack-extension-driven rupture scenario (industry-average crude-tower incident cost $50-150M per CSB and EPA major-accident records). Direct cost avoidance vs running with the LOSF unaddressed under FFS is harder to defend in dollars but obvious in regulatory and insurance posture.

Certifications required

Crew qualifications: