7 min read

Structural Weld Quality Trend Analysis: NCR Data to Action

How to use nonconformance reports and inspection records to identify systemic weld quality problems in a structural fab shop — Pareto analysis, root cause categories, and AISC audit expectations.

New to procedures? Start with WPS welding: what a welding procedure specification is, with a filled-in example.

A structural fab shop that generates NCR paperwork but never analyzes it is running a documentation system, not a quality management system. The difference matters — both for AISC certification and for the actual defect rates that determine rework cost, schedule impact, and long-term competitiveness.

Trend analysis converts your inspection records from a legal liability shield into an improvement engine. It's not complex statistical work. For most shops, a simple Pareto chart updated monthly reveals the two or three process problems that account for 70% of your rework hours. This article explains how to build that analysis and what to do when it identifies a systemic problem.

Rule library based on AWS D1.1:2025; verify against your governing edition — the AHJ or contract may specify 2020 or an earlier revision.

The Difference Between Tracking and Analysis

Most structural fab shops track NCRs. They have a log — paper or digital — that records each weld rejection, what was found, and how it was resolved. That's the minimum required for AISC certification and for demonstrating compliance with AWS D1.1 workmanship requirements.

Tracking answers: What happened?

Analysis answers: Why does it keep happening, and to which welds?

The tracking-only approach catches the individual defect. The analysis approach catches the systemic condition that keeps producing defects at the same joint, with the same process, or by the same welder. The rework cost of a single NCR is an annoyance. The cost of fifty NCRs on the same joint configuration over a year — each caught and repaired individually, without connecting the pattern — is a structural problem in your quality program.

Minimum Data Fields for Useful Analysis

An NCR that records only "undercut on W24 beam flange weld, repaired" is useless for trend analysis. You need the dimensions of the problem. At minimum, every NCR in a structural weld program should capture:

  • Defect type: undercut, porosity, incomplete fusion, crack, overlap, dimensional, profile violation, etc. Use a fixed pick list — not free text — so counts are meaningful.
  • Detection method: visual, MT, PT, UT, RT. This distinguishes surface defects found by CWI from subsurface defects found by NDE.
  • Weld process: SMAW, FCAW-G, GMAW, SAW, GTAW. Defects by process reveal consumable, parameter, or technique problems.
  • Joint type and configuration: CJP butt, CJP T-joint, PJP, fillet, etc. Certain joint geometries have characteristic failure modes.
  • Position: 1G/1F, 2G/2F, 3G/3F, 4G/4F, 6G. Positional defect rates reveal welder technique limitations or WPS parameter inadequacies for overhead or vertical work.
  • Welder ID: The most direct input for identifying individual welder performance issues.
  • WPS number: Connects defects to the procedure in use — useful when a WPS has parameters that are marginal for a given joint configuration.
  • Disposition: repair (re-weld), accept by engineering, replace. Track repair cycle count — a weld that required two repairs indicates a more serious underlying problem than one that passed on the first repair attempt.

These fields cost 5–10 minutes per NCR to populate. They return hours of debugging time when patterns emerge.

Pareto Analysis: Finding the Vital Few

The Pareto principle — roughly, that 80% of your problems come from 20% of your causes — applies to weld quality just as it applies to manufacturing defects. In a typical structural fab shop, the NCR data almost always produces a Pareto curve showing:

  • Two or three defect types that account for 60–70% of all NCRs
  • One or two welders who account for a disproportionate share of rejections
  • One or two joint configurations (often out-of-position welds or high-restraint joints) that generate most of the rework

Build this chart quarterly from your NCR log. A simple spreadsheet sorted by defect type, process, and welder produces an adequate Pareto view. The chart needs to be reviewed by someone with the authority to act on it — the QC manager or CWI lead — not just filed.

When incomplete fusion on CJP T-joints accounts for 40% of your NCRs, that's not random noise. That's a systemic signal pointing to a specific problem: joint fit-up tolerances, preheat inadequacy, WPS parameters for the joint geometry, or welder technique. See the incomplete fusion controls article for the technical factors.

Root Cause Categories for Structural Weld Defects

Once your Pareto identifies the dominant defect type, the root cause investigation follows one of four paths:

WPS parameter problem: The documented parameters for travel speed, amperage, voltage, or electrode extension produce marginal fusion for the specific joint geometry in production. This is common when a WPS was qualified on a flat plate test coupon but is used for vertical or overhead T-joints with restricted access. Solution: conduct additional PQR tests under actual production conditions, or add a WPS amendment with verified parameters for the problematic configuration.

Joint fit-up nonconformance: Root openings outside the qualified range, excessive hi-lo, or inadequate groove angle are driving fusion defects. CWI inspection records should reveal whether joints are routinely being accepted at the tolerance boundary — or whether fit-up deviations are present but not being caught. Solution: tighten the hold point at fit-up and verify dimensional data is being recorded per groove weld fit-up tolerances.

Welder technique problem: One or two welders account for a disproportionate share of a defect type. This is a training and qualification issue, not a WPS or fit-up problem. The corrective action is retraining, monitored production trials, or in persistent cases, restricting the welder's qualification scope until remediation is complete. Document the corrective action in the welder's qualification record.

Preheat and interpass temperature deviation: Hydrogen-related cracking and certain incomplete fusion modes correlate directly with preheat deviations. If your NCR data shows crack-type discontinuities clustering on a specific base metal or thickness range, check whether preheat verification records exist for those joints — and whether the measured temperatures met the WPS minimums. See the preheat verification methods guide for documentation best practices.

Connecting Defect Data to WPS and PQR Review

Persistent defects of the same type on the same WPS may indicate that the WPS parameters themselves need revision. Under AWS D1.1:2025, your ability to revise a tested WPS is constrained by the essential variable rules of Table 6.6 — changes within the non-essential variable range can be made without requalification. But some corrective actions require a new PQR.

If NCR data shows that a particular WPS consistently produces marginal incomplete fusion on 1-in. plate in the 3G position — and that position and thickness are within the qualified range on paper — the production data is telling you that the qualified range is at the boundary of what the procedure actually delivers. That's a reason to revise parameters and potentially run a production weld test, not a reason to keep accepting marginal welds or running more repairs.

Track the WPS number on each NCR specifically so you can identify which procedures are generating the most nonconformances. A WPS with a high NCR rate is a candidate for review regardless of whether individual welds are being accepted.

AISC Certification Audit Expectations

For shops seeking or maintaining AISC Standard for Steel Building Structures (AISC 303) certification, the certification body evaluator will review your NCR log as part of the QC system audit. What they're looking for:

Evidence that quality data is reviewed: The log should have a date-stamped quarterly or monthly summary review. A log that runs for a year without any analysis note is a gap finding.

Corrective action records: For recurring defects — same defect type appearing three or more times on the same process or configuration — there should be a documented corrective action investigation and a resolution record. The corrective action doesn't need to be elaborate, but it must be specific. "Remind welders to maintain preheat" is not corrective action. "Revised WPS-14 to increase minimum preheat on A36 over 1.5 in. from 50°F to 150°F per Table 5.11; CWI to verify with contact thermocouple at fit-up hold point" is corrective action.

Trend visibility at the management level: The QC manager or CWI lead should be able to show the auditor the top two or three defect types in the past 12 months without having to sort through raw NCR paper. If you can't answer "what were your most common weld rejections last year?" in under two minutes, your quality system isn't using the data it's generating.

For the full picture of common weld quality program deficiencies found during AISC audits, see the AISC fab cert audit failures article.

Closing the Loop: From Data to Procedure Update

The end state of a mature weld quality trend analysis program is a feedback loop: inspection data → trend analysis → root cause identification → corrective action → WPS/training/fit-up procedure update → verification that the corrective action reduced the defect rate.

This loop runs on data. It requires that NCRs be captured consistently, categorized consistently, and reviewed regularly by someone with authority to change procedures. The investment is modest compared to the rework cost of systemic, undetected problems.

A structural fab shop ready to build this system — from WPS documentation through welder qualification tracking and NCR-linked inspection records — can find the tools to do it at WPS Welding. Managing your quality data in the same platform as your WPS library makes the connection between procedure documentation and production quality visible rather than theoretical.

FAQ

Frequently asked.

How many NCRs should trigger a formal trend analysis in a structural fab shop?
There's no universal threshold in AWS D1.1 or AISC certification requirements, but most QC plans specify a review when a defect type exceeds a defined rate — commonly 2% of inspected joints for a given defect type. The more useful trigger is a repeated defect on the same joint configuration, process, or welder, which indicates a systemic problem rather than random variation.
What categories of data should be tracked on each NCR?
At minimum: joint type (CJP, PJP, fillet), weld process (SMAW, FCAW-G, SAW), welder ID, joint configuration (T-joint, butt, corner), position, defect type, detection method (visual, MT, PT, UT, RT), and disposition (repair, accept-as-is, replace). Tracking shift and inspector can reveal inspection variability issues separately from production quality issues.
Does AWS D1.1 require a corrective action program for repeated defects?
AWS D1.1 itself does not mandate a formal corrective action program, but it does require that rejectable discontinuities be repaired and that repairs meet the same acceptance criteria as original welds. AISC Chapter N and certification body requirements do require a written quality control plan that addresses how nonconformances are identified, documented, dispositioned, and prevented from recurrence — which implies a corrective action process.
What does an AISC auditor look for in terms of weld quality trend data?
AISC certification auditors typically review the NCR log and ask whether the QC manager can identify the top defect types and what corrective action was taken. A log that lists NCRs with no analysis, no recurring issue identification, and no corrective action is an audit finding. The auditor wants to see that quality data drives process improvement, not just documentation.

More from the blog