PFMEA Welding Example (Robotic MIG/MAG)
Process overview
The scope is a welded structural steel subassembly (e.g. an automotive cross-member or an industrial frame) produced on a two-station robotic cell with MIG/MAG welding, solid wire, Ar/CO₂ shielding gas, and post-weld UT on Class A joints. Material is S355 structural steel; joints include fillet and groove welds, with one safety-critical butt joint requiring 100 % UT. The PFMEA covers fixturing through final dimensional verification; weld procedure qualification (WPQR) and welder certification are upstream quality-system controls referenced but not re-rated here.
Process flow
- Part loading and fixture clamping
- Tack welding (manual or robotic)
- Robot weld execution — program selection per part variant
- Wire feed and shielding gas delivery
- Post-weld cooling on the fixture
- Visual inspection per ISO 5817 / customer standard
- NDT (UT, MT or RT) on Class A joints
- Deburr / grinding of weld toe where required
- Final dimensional check on CMM or gauge
Common failure modes
- Lack of fusion — sidewall or root; the dominant safety failure mode and the hardest to detect.
- Porosity — from gas-flow loss, contaminated wire or surface contamination.
- Undercut — visible defect but easily missed under poor lighting; a fatigue initiation site.
- Cracking — solidification or hydrogen-induced (HIC); particularly on higher-carbon steels.
- Lack of penetration — wrong parameters, magnetic blow, or fitup gap.
- Distortion — predictable from heat input; only controlled by sequence and fixturing.
- Wrong weld program — the program selection error on multi-variant lines is the most common cause of out-of-spec welds.
Representative PFMEA table
| Process Step | Failure Mode | Effect | Cause | Prevention | Detection | S | O | D | AP |
|---|---|---|---|---|---|---|---|---|---|
| Part loading & fixturing | Wrong part orientation in fixture | Weld in wrong location → rework or scrap | Symmetric fixture allows asymmetric loading | Asymmetric fixture pins / Poka-yoke | First-piece visual + robot vision check | 7 | 4 | 3 | Medium |
| Tack welding | Missing tack weld | Distortion or fitup gap on main weld | Operator skip; tack program error | Mandatory tack-count verification by robot vision | Pre-weld visual at first robot station | 7 | 3 | 4 | Medium |
| Robot weld execution | Wrong weld program selected | Wrong current / voltage / travel speed → defective weld | Program selector error or mismatch with fixture | Program selection tied to fixture RFID | Weld data acquisition system with parameter limits | 8 | 3 | 3 | High |
| Wire feed | Wire feed slip / arc loss at start | Lack of fusion at weld start | Worn drive rolls or liner blockage | Drive roll PM + liner inspection schedule | Arc-on monitoring with auto-stop on dropout | 8 | 4 | 4 | High |
| Shielding gas | Insufficient gas flow | Porosity in weld; loss of toughness | Regulator drift or hose leak | Daily gas flow meter check + leak audit | Visual porosity inspection + post-weld MT | 8 | 3 | 5 | High |
| Robot welding | Arc deflection (magnetic blow) | Lack of penetration; off-axis bead | Magnetized clamps / nearby DC return | Demagnetize fixtures on schedule; manage ground placement | UT sampling on root pass for critical joints | 8 | 3 | 5 | High |
| Post-weld cooling | Forced cooling on hardenable steel | Quench cracking; brittle HAZ | Operator uses fan to speed cycle | Cooling procedure WI + temperature-managed cycle time | Magnetic particle inspection (MT) on Class A joints | 9 | 2 | 4 | High |
| Visual inspection | Undercut > 0.5 mm missed | Stress concentration → fatigue failure in service | Inspector fatigue or insufficient lighting | Inspector rotation + standardized lighting station | Independent sample re-inspection | 7 | 4 | 5 | High |
| NDT (UT / MT / RT) | Sub-surface lack of fusion missed | Field failure under load on safety-critical joint | NDT operator skill / calibration block missing | ASNT Level II certification + calibration block per shift | Independent NDT audit on 5 % of lots | 9 | 2 | 4 | High |
| Final dimensional check | Weld-induced distortion exceeds GD&T | Assembly interference downstream | Fixture wear; weld sequence not optimized | Fixture verification schedule; weld sequence FEA | CMM check of welded subassembly | 7 | 3 | 3 | Medium |
Example failure chains
| Trigger | Failure chain | Customer impact |
|---|---|---|
| Gas regulator drift overnight; shielding gas flow drops 30 % | Porosity in first 20 welds → visual inspection catches surface porosity but not sub-surface → fatigue initiation on welds shipped before re-check | Field failure on cyclic load, weeks to years later |
| Operator selects program for variant B on a variant A fixture | Wrong current/voltage profile → lack of penetration on Class A joint → UT sampling catches one in 20 → unsampled units already shipped | Customer recall on suspect lot, full UT re-inspection cost |
| Drive roll wear → wire feed slip at arc start | Lack of fusion at start of weld → buried inside the bead → not visible to AOI or visual → only UT finds it | Field failure or audit non-conformance |
Current controls
- Prevention: RFID-bound program selection, drive-roll PM, daily gas flow check, fixture demagnetization schedule, welder certification matrix, WPQR.
- Detection: weld data acquisition with parameter limits, arc-on monitoring, post-weld visual per ISO 5817, MT or UT on Class A joints, 5 % independent NDT audit, CMM dimensional verification.
Recommended actions
- Install closed-loop gas flow monitoring with reject on drop below threshold.
- Move from RFID program selection to fixture-keyed mechanical lockout to eliminate human selection error entirely.
- Increase NDT sampling on Class A joints from 100 % visual + sampling UT to 100 % UT during first 30 days after any process change.
- Implement closed-loop arc monitoring with auto-stop on first dropout, not after the weld completes.
Action Priority discussion
Welding PFMEAs are dominated by AP High rows because Detection on sub-surface defects is fundamentally weak. Even with NDT, Detection is rarely better than 4 for sub-surface lack of fusion. The AIAG-VDA AP table correctly promotes these rows to High; teams using legacy RPN often miss this because RPN multiplies rather than ranking by Severity-first logic. See RPN vs Action Priority for why this matters specifically in welding.
Common mistakes in welding PFMEAs
- Scoring Detection 2 for "100 % visual". Visual cannot see sub-surface defects; realistic Detection is 5–6.
- Ignoring program selection error on multi-variant cells. It is the dominant cause of out-of-spec welds on real lines.
- Treating gas flow as a setup parameter, not a process variable. Daily flow check belongs in the PFMEA.
- Failing to distinguish Class A safety-critical joints from Class B / C welds with the same row. Severity differs by joint class.
- Not linking the PFMEA to the WPS / WPQR. Auditors will look for it.
Audit considerations
Welding audits — whether ISO 3834, IATF 16949 or customer-specific — look for: WPS / WPQR alignment with PFMEA, NDT operator certification (typically ASNT Level II or ISO 9712), calibration block control, shielding gas verification records, and traceability of welds to operator and machine. Each High-AP row in this PFMEA should be traceable to a specific control in the Control Plan and a specific WPS line.
PFMEA — Welding in the Qhubio knowledge graph
How PFMEA — Welding connects to other FMEA concepts, standards, examples and software.
Frequently asked questions
Learning path
Three tiers, automatically derived from the Qhubio knowledge graph. The tier containing this guide is highlighted.
Further reading
Curated next steps — methodology guides, worked examples, and the relevant tool.
Qhubio FMEA solutions
Move from theory to a working FMEA. Same methodology, three entry points depending on how you want to start.
Generate a structured FMEA in minutes
Qhubio applies the AIAG-VDA methodology automatically — no Excel formulas, no inconsistent rating scales, no scattered spreadsheets.
Generate your first FMEAHelp us improve Qhubio. Analytics help us understand which FMEA features are useful, which pages need improvement, and where users encounter problems. No advertising. No selling personal data. See our Cookie Policy.