ISO 5817 vs ASME & AWS: Weld Acceptance Criteria Compared

Fabricators selling into both European and North American markets hit the same wall: a weld sentenced “acceptable, ISO 5817-C” lands in front of an ASME inspector, or an ASME-coded vessel weld is audited by a customer who thinks in quality levels. The two systems do not map onto each other — and pretending they do is how dual-compliance shops accumulate findings.

This article compares how the EN ISO and ASME/AWS worlds structure weld acceptance: philosophy, where the criteria actually live, and how to run one production system that satisfies both.

Two architectures for the same question

The structural difference matters more than any individual limit:

• ISO 5817 is a horizontal standard — one imperfection framework reused across industries, with strictness set by choosing B, C or D. The full mechanics are covered in our ISO 5817 acceptance criteria guide.
• ASME/AWS criteria are vertical — each code embeds its own acceptance tables, written around the failure modes of its product class: pressure boundary integrity for ASME B31.3 piping, static vs cyclic loading for AWS D1.1 structures, pipeline girth welds for API 1104.

The misconception that costs audits: Section IX

ASME Section IX does not contain acceptance criteria for production welds. It qualifies procedures and people — the WPS/PQR/WPQ chain (digital workflow here). Production acceptance comes from the construction code the part is built to.

The ISO world has the same split, which makes the mapping clean:

Where the criteria actually diverge

Without reproducing copyrighted tables, the practical divergences fall into four buckets:

1. How strictness is selected

ISO: the application standard maps to a level — EN 1090-2 execution classes map EXC1→D up to EXC4→B (details in the B vs C vs D comparison). ASME: strictness follows service category (e.g. normal vs severe cyclic in B31.3) and the examination performed — criteria for radiographed joints differ from visually examined ones.

2. What is measured

ISO 5817 limits are predominantly dimensional formulas scaled to thickness (h ≤ 0.1t style). ASME code limits mix absolute dimensions, fractions of wall thickness, and indication-length rules tied to the NDT method. The same undercut can therefore pass one framework and fail the other depending on t and the examination performed.

3. Zero-tolerance lists largely agree

Both worlds reject cracks and lack of fusion outright in essentially all configurations — planar defects get no dimensional debate anywhere (see why planar defects are different). Convergence here is no accident: it’s fracture mechanics.

4. Inspection extent

EN 1090/ISO 3834 set NDT percentages via execution class and joint type; ASME codes set examination extent per service category (100% RT vs spot RT changes both the inspection and the applicable acceptance criteria — a B31.3 peculiarity that surprises ISO-trained inspectors).

Running one shop for both worlds

What works in dual-certified fabricators:

1. One imperfection vocabulary internally — ISO 6520-1 codes, mapped to code terminology at the sentencing record, so NCR statistics aggregate across markets.
2. A mapping matrix per imperfection type: ISO 5817 level limit vs applicable ASME/AWS limit, sentencing always to the stricter of the two for dual-destination parts.
3. Records that cite both criteria — “accepted per ISO 5817-C and B31.3 Table 341.3.2” closes the audit question before it’s asked. This is where digital weld records beat paper: the same evidence object serves both frameworks.
4. Process monitoring as the common layer: imperfection generation doesn’t care which code sentences it. Thermal and high-speed visual monitoring of every weld reduces the defect population reaching either sentencing gate, and the per-weld evidence satisfies both ISO 3834 traceability and ASME QC documentation expectations.

For the defect-by-defect physics underneath both systems, see the welding defects guide.