First article inspection on a large weldment

An FAI report that arrives as forty numbers with no stated datums, no equipment identification and no uncertainty statement is not evidence. It is an assertion formatted as a table, and you cannot reconcile it against your drawing without redoing the measurement.

On a four-metre machine frame that redo costs a week and a crane. The way to avoid it is to specify the inspection before the part is made, not to evaluate the report after. The first article is the gate in our guide to machine frame fabrication sourcing, before the series is released.

Hold points, not one inspection at the end

A single final inspection on a large weldment is a bad design of process. By the time the part is finished, every earlier mistake has been buried under expensive operations, and the only remedies are scrap or rework on a part that has already absorbed most of its cost.

Four hold points cover a machine frame properly:

1. Fit-up, before welding. The assembly is tacked and jigged. Check overall dimensions, squareness, diagonals and joint gaps. This is the cheapest correction opportunity in the entire process — a tack can be broken and reset in an hour. Photographs at fit-up cost nothing and settle most later disputes in one email.

2. After welding, before stress relief. Shows what the weld sequence actually did. The question being answered is whether the machining allowance survived the distortion. If it did not, you find out now rather than when the cutter runs out of material. The sequencing levers are in controlling distortion on large weldments.

3. After stress relief, before finish machining. Shows what relaxation added, and confirms the part is worth putting on an expensive machine. Also the point at which the furnace chart should be reviewed, as stress relieving welded machine bases sets out.

4. Final, after machining. The full dimensional report against the drawing, plus coating and documentation checks.

Booking these into the PO costs a small amount of inspection time and removes the two expensive failure modes: discovering a fabrication error after machining, and discovering a machining error after coating.

Measuring a part you cannot put on a table

A four-metre frame will not go on most surface plates, and a CMM with that envelope is a rare and expensive machine. Three practical approaches, in ascending order of capability:

Straightedge, feeler gauges and precision level. Old, cheap, and genuinely accurate for flatness and straightness on long surfaces when done carefully. A precision level walked across a machined plane in a grid resolves flatness to a few microns per metre. Slow, and entirely adequate for many frames.

Laser tracker. The standard tool for large-part metrology. Volumetric accuracy in the tens of microns over several metres, measures features anywhere in the working volume, and produces a report in your coordinate system. This is what most competent heavy-fab shops use or hire for large frames.

Photogrammetry or laser scanning. Fast, dense point clouds, excellent for capturing whole-surface form and for comparing as-built to CAD. Less accurate than a tracker for discrete features, very good for flatness maps and for showing where a surface deviates rather than just by how much.

Specify which is acceptable. A frame with 0.02 mm flatness callouts inspected with a tape measure and a straightedge has not been inspected.

Measurement uncertainty, which belongs in the report

Every measurement has an uncertainty, and on large parts it is not small. Thermal state alone can dominate: steel moves roughly 12 μm per metre per degree Celsius, so a four-metre part measured at 28 °C and specified at 20 °C is nominally 384 μm longer than nominal before any error is considered.

Require three things in the report:

• Temperature at measurement, of the part and not just the room, and whether a correction was applied.
• Equipment identification and calibration status — serial number and last calibration date.
• A statement of measurement uncertainty for the reported characteristics.

If the uncertainty is a meaningful fraction of the tolerance, the tolerance is not really being verified. That is worth knowing at FAI rather than after a disagreement, and it occasionally reveals that a tolerance on the drawing was never measurable in the first place — a problem better solved by revising the drawing, as tolerance stack-up on fabricated machine structures discusses.

Datums and support, which decide whether the numbers mean anything

Two requirements, and both are commonly missed.

Measure from the drawing’s datums. A report referenced to features the shop found convenient cannot be compared to your drawing. Datum targets are what make this practical on a large machined face — the datum plane is established from three defined points rather than from an entire surface nobody can simulate. The scheme is set out in GD&T and datum strategy on welded structures.

Define the support scheme, and use the same one for machining and inspection. A large weldment sags under its own weight. Supported on three points it has one shape; on eight it has another. If the part is machined on one support arrangement and inspected on a different one, the two sets of numbers describe two different objects, and the disagreement is nobody’s fault.

State it explicitly: part to be supported at the six points marked on sheet 3 during both final machining and final inspection.

What the report must contain

A usable FAI package for a machine frame:

1. Part number, revision, drawing revision, serial or heat number
2. Ballooned drawing — every characteristic numbered, every number reported
3. Measured values against nominal and tolerance, with pass/fail
4. Datum reference frame used, matching the drawing
5. Support scheme and part orientation during measurement
6. Equipment list with calibration dates
7. Temperature of part and environment, and correction applied
8. Measurement uncertainty statement
9. Material certificates (EN 10204 3.1) traceable to heat numbers
10. Welding records — WPS, welder qualifications for the joints made
11. NDT reports with operator qualification, per the critical weld table
12. Stress relief furnace chart
13. Coating records — blast grade photographs, DFT readings, adhesion test
14. Photographs at each hold point
15. Non-conformance reports and approved dispositions, if any

Item 15 matters more than its position suggests. A package with no NCRs on a first large weldment is unusual, and an FAI that reports nothing went wrong is either a very good shop or an incomplete record. Ask which.

Writing it into the PO

Inspection. Hold points at fit-up, post-weld, post-stress-relief and final. Supplier to notify 5 working days before each. Measurements referenced to datums A, B and C per drawing, part supported at the six points on sheet 3 for both final machining and final inspection. Large-envelope characteristics by laser tracker or equivalent; equipment calibration certificates to be supplied. Report to state part temperature and measurement uncertainty. Full FAI package per the list in Annex A to be accepted before shipment.

The last clause — acceptance before shipment — is the one that gives the requirement teeth. Without it, the package arrives with the crate and any problem it reveals is a problem you own. Where this sits in the schedule is shown in the machine frame sourcing timeline.

If you want help defining the hold points and the FAI content for a frame before it goes out to quote, send the drawing through — the review comes back within 24 business hours.