Direct answer: Weld acceptance goes wrong most often when two different things are written as one clause. Bead profile, width and the stop position can be measured on the finished part. Fusion depth and root fusion sit inside the joint and cannot be inferred from the outside, so they belong to a qualified procedure and to inspection. Written as separate gates, both become checkable; written together, neither does.
Who this is for: This guide is written for quality and manufacturing engineers writing acceptance conditions for automated welding equipment, where the same document has to cover what an operator can check and what only a laboratory can establish.
Scope: It covers how visible profile checks, hidden fusion characteristics and procedure qualification belong to different gates, and what each judgement still owes in evidence. It does not cover inspection technique selection, weld design, or the choice of consumables.
Two different things written as one clause
Most disputes over weld acceptance at handover are not about welding at all. They are about a sentence in an acceptance document that promised a good weld without saying who checks what, when, and by what means. By the time the argument starts, the equipment is installed and the leverage has moved.
The split that makes weld acceptance workable is simple to state. One group of characteristics is on the outside of the joint and can be observed and measured on a finished part. Another group is inside the joint and cannot be observed at all, however carefully anyone looks. Writing them into one clause guarantees that at least one of them is unenforceable.
The reference footage happens to contain both halves in the right order. A robot welds a branch tube into a main tube held in a rotating chuck, and the same clip then moves to close-ups of the finished joint with a caliper against it. What the caliper reaches is exactly the boundary this article is about.
What the profile can tell you
Reinforcement, bead width, uniformity along the joint, undercut at the toes and where the arc stopped are all on the surface. They can be checked visually, measured with simple gauges, and judged against a stated level, which is why they are the natural content of a line-side gate.
According to ISO 5817:2023, quality levels for imperfections in fusion-welded joints are defined and assessed on the joint itself rather than assumed from the equipment or the operator, and stating the level being applied is what turns a vague requirement into a checkable one. An acceptance document that names the level has said something; one that asks for a good weld has not.
Automated welding gives the surface check an extra role. Because the same program runs with the same fixturing, profile consistency is unusually high, and that consistency makes appearance a sensitive process monitor. When the shape starts to drift, fixturing, incoming parts or parameters have usually drifted first, and the visible change is the early warning.
What stays inside the joint
Fusion depth, root fusion and internal discontinuities are the characteristics that carry most of the load and show none of themselves. No amount of surface tidiness demonstrates them, and experienced judgement from the outside is an estimate rather than a measurement. This is the half that acceptance documents most often leave implicit.
It is also the half that gets settled once rather than continuously. According to ISO 15614-1:2017, a welding procedure is qualified by test, and the qualification defines the range over which the tested procedure remains valid. Production then works inside that range, with verification appropriate to the criticality of the product rather than a repeat of the qualification on every part.
Joint preparation belongs to the same conversation, because it determines what the process has to achieve before any parameter is chosen. According to ISO 9692-1:2013, joint preparation types are specified for the process and material combination being used, which is why a change of preparation is a change to the basis of the qualification and not a detail to be handled on the shop floor.
Decision table: which gate each requirement belongs to
The table maps common acceptance requirements to the gate that can actually close them, and to the evidence each one needs.
| Requirement as usually written | Gate that can close it | Evidence it needs |
|---|---|---|
| The weld shall be free from visible defects | Line-side visual gate on the finished part | A named quality level, the features checked, who checks and at what frequency |
| The weld shall be fully fused at the root | Procedure qualification, plus production verification by inspection | Qualified procedure with its validity range, and the verification method and extent |
| Welds shall be consistent across production | Line-side profile monitoring against the qualified condition | Reference profile, drift criteria and the action taken when the shape moves |
| Equipment shall produce acceptable welds | Neither gate as written; it has to be split before it can be tested | Separate visible and internal clauses, each with its own method and acceptance rule |
Procedure tests are the bridge between the two gates
The reason a split document works is that qualification connects the two halves. A qualified procedure establishes that a given combination of joint, material, position and parameters produces the internal result required, and it states the range over which that conclusion holds.
Once that exists, the line-side gate has something to compare against. Visible profile is no longer a proxy for internal quality, which it cannot be; it becomes evidence that production is still running inside the qualified condition, which it can be. That is a much narrower claim and a far more defensible one.
According to ISO 3834-2:2021, comprehensive quality requirements are met through documented arrangements covering the whole fabrication rather than through inspection of the finished item alone, which is another way of saying the same thing. In practice the documents that survive handover are the ones that state, for each requirement, which gate closes it and what evidence is retained. EVST writes acceptance conditions in that shape for exactly this reason.
What the footage proves and what it does not
The reference footage proves a visible process. A robot welds a branch tube into a main tube while a chuck rotates the part; the finished saddle joint shows a regular bead; and a caliper is used against the completed weld. Those are observable facts about the operation and the surface of the joint.
It proves nothing about fusion depth, root condition, internal discontinuities or the acceptance result, and it was never capable of doing so. That is not a limitation of this particular clip. It is the point the article is making, and it is why the two gates exist.
Keeping the two apart in writing is what makes a weld acceptance document testable. Each requirement names its gate, its method and its evidence, and both parties can tell before installation whether a requirement can be closed at all.
What to put in the acceptance document
State the joint and the material: geometry, preparation, thickness, position and the service the joint sees. Then state the visible gate: the quality level being applied, the features checked, the method, who performs the check and how often. Related process context is worth referencing so the boundary is clear, such as multi-pass welding process planning where several passes fill one joint, or robotic welding and cutting path boundaries where the path itself constrains what is achievable.
Then state the internal gate separately: the procedure qualification required, its validity range, the verification method and extent during production, and what happens when something falls outside. Add the records to be retained for each gate. Written this way the same document supports a line-side selection between conforming and non-conforming parts and a defensible position at handover, and equipment such as automatic welding system solutions can be assessed against conditions that both sides can actually test.
Frequently asked questions
Can a good-looking weld be rejected?
Yes. Surface appearance and internal fusion are independent characteristics. A joint can be neat outside and inadequate inside, which is why the internal gate is qualified separately and verified by an appropriate method rather than by inspection of the surface.
Do we need to inspect every weld internally?
Rarely. The usual arrangement qualifies the procedure by test, then verifies during production at an extent matched to the criticality of the product. The acceptance document should state that extent explicitly rather than leaving it to be negotiated later.
Why is profile still worth measuring on automated welding?
Because it is a sensitive process monitor. The same program and fixturing produce highly consistent shapes, so a change in profile usually means fixturing, incoming parts or parameters have already changed. It is early warning, not proof of internal quality.
Does the footage show whether these welds are acceptable?
No. It shows a joint being welded and the finished bead being measured. Acceptance depends on the quality level applied, the qualified procedure and the verification result, none of which are visible in the picture.
Project inputs for an application review
Send the following and the acceptance conditions can be written as gates that both sides can test:
- the joint geometry, preparation, material, thickness and welding position
- the service condition the joint sees, and the criticality of the product
- the visible quality level you intend to apply, and who performs the check
- the verification method and extent you expect during production
- the records each party expects to retain, and for how long
If you are writing acceptance conditions for automated welding equipment, send the joint geometry and preparation, material and thickness, the service condition, the visible quality level you intend to apply, and the verification method and extent you expect. The conditions can then be split into a line-side gate and a qualified-procedure gate, each with its own evidence. Related reading: multi-pass welding process planning, robotic welding and cutting path boundaries.