Robotic Welding Cell Acceptance: What to Record
Direct answer: At acceptance the question is not whether the cell can weld but whether the result repeats and can be traced. Record how the part is located and clamped and in what order; arc start points, which have to reproduce; segment count and tie-in positions on long seams; positioner return accuracy after each index; and crater positions. Penetration and the verdict come from the standard and the inspection report.
Who this is for: Written for engineers and buyers preparing to accept a robotic welding cell and the record that has to come with it.
Scope: This EVST guide covers what belongs in a robotic welding cell acceptance record: how the part is located and clamped, arc start reproducibility, segment tie-in positions on long seams, positioner return accuracy and crater positions. It does not quote penetration, quality levels or pass or fail results; those come from the acceptance standard and the inspection report.
Acceptance is not a demonstration
Whether a cell can produce the seam is settled at proposal stage, on test pieces. By acceptance that question is behind you, and the useful ones are whether the result repeats and whether a problem can be traced back to something written down.
The reference footage shows a single robot working with a headstock and tailstock positioner on a thick plate structural part, clamps fixed along the fixture. In that arrangement repeatability comes from a short list of items.
The record should mirror that list one item at a time. A single line saying the cell was accepted is not a record; it is a signature on the absence of one.
The robotic welding cell acceptance criteria worth writing down
Clamping comes first: what locates the part, where the clamps sit, and in what order they are applied. Those three decide whether the part lands in the same place each cycle, which is the condition every taught program depends on.
Arc start points come second. Each segment has to start from a position that reproduces, or batches cannot be compared and a defect cannot be located to a place on the part.
Both have to be written as something that can be followed rather than described. The value of an acceptance record is that a different person on a different shift reaches the same result from it.
An EVST acceptance list is built onto whatever record format is already in use, so it adds items rather than replacing a working document.
According to ISO 3834-2:2021, comprehensive quality requirements for fusion welding include documented arrangements for procedures, personnel and equipment, which is the framework this record sits inside.
Long seams are recorded by segment
Long seams are usually welded in segments. How many segments, where each begins and ends, and where the tie-ins fall all belong in the record; a single line describing the seam is not enough.
Tie-in position is what later inspection and any repair are located from. Without it a defect report starts from the beginning of the seam every time.
If the segment count changes in production, the record has to change with it. Otherwise the document and the process drift apart quietly, which is the failure mode acceptance records are supposed to prevent.
| Record item | What it protects | How it is verified |
|---|---|---|
| Locating and clamping, with order | That the part lands where the program assumes | Repeat set-ups of the same part, checked against the same reference |
| Arc start points per segment | Comparability between batches, and defect location | Repeat runs starting from the recorded positions |
| Segment count and tie-in positions | Inspection and repair can be located on the part | Marked drawing or sketch kept with the record |
| Positioner return accuracy after indexing | That the second face is welded to the same geometry | Repeated index and return, measured against a reference feature |
| Crater positions and their order | The traceable boundary of each segment | Recorded with the segment they close |
Repositioning after indexing is the item most often skipped
When the part is turned, it has to be located again. The return accuracy of the positioner is what makes the second face land on the geometry the program expects.
It gets skipped because nothing looks wrong while it works. The consequence appears later as seams that are acceptable on one face and marginal on the other, with no record to show when the difference started.
Each changeover repeated is another entry to keep. In the reference footage the part indexes and welding continues on a new face, and each of those transitions is a record point.
According to ISO 9283:1998, pose repeatability is measured under stated load and speed, so a return accuracy statement has to name the part mass and the conditions it was taken under.
What the record does not contain
Penetration, appearance grade and the pass or fail verdict do not belong in the acceptance record. They belong to the inspection report, which is a separate document answering a separate question.
According to ISO 5817:2023, weld imperfections are graded into quality levels, and which level applies to your joints is a specification decision taken before the cell is built.
According to ISO 17637:2016, visual testing of fusion welds is carried out under stated conditions of access, lighting and surface condition, which is why a visual result is only meaningful with those conditions recorded.
Keeping the two documents separate is what makes it possible to work backwards from a result to the process that produced it, instead of re-running trials to find out what changed. In practice an EVST acceptance review starts from the drawing, the applicable standard and whatever record format is already in use.
Every application review opens from the same project inputs – drawing, standard and current method – so selection rests on evidence and acceptance is agreed against your own parts rather than a datasheet.
Frequently asked questions
Is an acceptance record the same as an inspection report?
No. The acceptance record describes how the result is produced – locating, clamping, arc starts, segments, indexing. The inspection report says whether the produced weld meets the specified quality level. Keeping them separate is what makes a later problem traceable.
Why record arc start positions?
Because they make batches comparable and give inspection and repair a location on the part. If each run starts somewhere slightly different, a defect cannot be tied to a position or a cause.
What is usually missed at acceptance?
Positioner return accuracy after indexing. Nothing looks wrong while it works, and when it drifts the effect shows up as one face being consistently worse than the other with no record of when it began.
What should we send to get a useful answer?
The part drawing, the acceptance standard and quality level you work to, and any record format already in use so the list can be built onto it rather than beside it.
Project inputs for an application review
To have the record list built against your own part and standard, send:
- part drawing with joint details and seam lengths
- the acceptance standard and quality level applied
- the current record or inspection format in use
- how the part is located, clamped and indexed today
Send the drawing and the standard you work to and the record list can be built against them. Related reading: robot welding workstation, self-learning 3D vision welding.