Direct answer: On medium-thick plate a single pass demands a heat input whose cost in grain growth and distortion exceeds the cost of several passes. Multi-pass welding therefore starts at the groove: angle, root face and root gap set the cross-section, and the cross-section sets the minimum pass count. Bead placement then drives reinforcement, inter-run profile and the risk of lack of fusion.
Who this is for: This guide is written for welding and manufacturing engineers planning robotic multi-pass welding on medium-thick plate, where a single pass is not an option and the pass plan is often left to the programmer.
Scope: It covers how groove geometry determines layer and bead planning, why bead placement drives both appearance and fusion, and what robotic welding adds to the problem. It does not cover procedure qualification testing itself, consumable selection, or the design of the structure being welded.
Why one pass does not finish a thick joint
As thickness increases, the heat input required to fuse the full section in one pass grows to a level that is difficult to justify. The consequences are metallurgical and dimensional at the same time: coarser grain structure, a wider heat-affected zone, and more distortion to correct afterwards.
Filling the joint in layers keeps the heat input of each pass within a workable range. It also introduces useful effects: each pass preheats the region for the next, and subsequent passes temper part of the material deposited before them.
The macro evidence is straightforward to read. A multi-pass weld does not present as one uniform band but as a set of distinct beads stacked in sequence, and the boundaries between them are visible on the finished surface.
In practice the pass plan is where robotic multi-pass welding either becomes repeatable or stays dependent on the programmer. According to ISO 15614-1:2017, welding procedures are qualified by test rather than by assertion, which is the mechanism that turns a pass plan into a documented procedure.
The groove sets the cross-section, the cross-section sets the passes
Groove angle, root face and root gap together determine the area that has to be filled. Divide that area by what a single pass can reasonably deposit and you have the lower bound on pass count. That is why the planning sequence starts at the groove and not at the parameter set.
Groove geometry also governs accessibility. Too narrow and the torch cannot enter, or can only enter at an unfavourable angle. Too wide and the volume to fill grows disproportionately, taking cycle time and distortion with it. ISO 9692-1:2013 covers types of joint preparation and provides a common vocabulary for that discussion.
Both directions have real costs, so the balance belongs in the design phase rather than in production. A groove chosen for ease of cutting can quietly commit the cell to significantly more filling than a slightly different preparation would have required.
According to ISO 9692-1:2013, joint preparation types are classified explicitly, which gives fabricator and integrator a shared vocabulary instead of an argument about what a bevel should look like. EVST starts multi-pass welding discussions from that classification.
Bead placement drives appearance and fusion
Where each pass is placed relative to the one before it determines whether reinforcement is even, whether grooves form between runs, and whether the boundary between layers is properly fused. These are related but distinct outcomes, and only the first is visible from outside.
Sequence within a layer also affects distortion. Different bead orders produce different angular distortion and different longitudinal shrinkage on the same joint, and on structural work that difference is often larger than the effect of parameter adjustment.
Interpass temperature belongs in the same plan. It influences both metallurgical properties and the formation of the next bead, which makes it a parameter to be specified and controlled rather than a result to be explained afterwards.
Decision table: what the joint justifies
The table maps what plate thickness and joint type can justify on their own against the evidence a supplier still has to produce.
| Evidence from the joint | Planning it justifies | Evidence you still owe |
|---|---|---|
| Consistent machined or thermally cut preparation, stable fit-up | A fixed layer and bead plan taught once and reused across the run | Measured fit-up distribution, and the pass plan verified against the qualified procedure |
| Manually prepared bevels with visible variation | Seam finding, or tighter control of the preparation method, decided before cell design | The actual variation range, since it decides whether sensing is required or optional |
| Thickness or joint type varies across the part mix | A pass plan per joint class rather than one plan applied to everything | Joint classes defined, with the acceptance level for each |
| Acceptance level not yet agreed | No final procedure; appearance and volumetric requirements separated in the discussion | Agreed quality level and the inspection method that will demonstrate it |
Groove consistency is the variable robots expose
Manual welding absorbs a certain amount of preparation variation because the welder observes the joint and adapts continuously. A robot executing a taught path does not, which converts preparation variation from a tolerable nuisance into a process input that has to be controlled or measured.
That leads to a decision that belongs early in the project: either tighten the preparation method so that variation stays inside the taught plan, or add seam finding or tracking so the cell can accommodate it. Both are legitimate; choosing neither is not.
The information needed to decide is the actual distribution of fit-up and bevel dimensions from current production. In the absence of that data, the decision is usually deferred to commissioning, which is the most expensive place to take it.
Appearance is not penetration
A macro view of a finished multi-pass weld demonstrates bead sequence, overlap and surface regularity. It does not demonstrate penetration, fusion between layers, internal soundness or mechanical properties, and it should not be presented as if it did.
ISO 5817:2023 defines quality levels for imperfections and gives both parties a shared vocabulary for what is acceptable. ISO 15614-1:2017 covers procedure qualification, which is the mechanism by which internal results are actually demonstrated.
Keeping the two categories separate is the single most useful discipline in this area. Visible evidence supports process direction; volumetric and mechanical claims require the corresponding examination.
Dimensional expectations belong in the specification
Multi-pass welding on structural work produces measurable shrinkage and angular distortion. Where dimensional expectations are left implicit, they are usually discovered at assembly, and the correction is attributed to welding rather than to the absence of an agreed tolerance.
ISO 13920:1996 provides general tolerances for welded constructions and is a practical starting point for that agreement. Referencing a class explicitly is more useful than describing the requirement in adjectives.
Where the structure has tighter local requirements, stating them separately keeps the general tolerance workable rather than forcing the whole assembly to the tightest case.
According to ISO 13920:1996, general tolerances for welded constructions are stated as classes, so quoting a class is more useful than describing the requirement in adjectives. EVST asks which class applies before the fixture concept is drawn.
What to put in the capital request
State plate thickness and material, joint type and the groove preparation you intend to use, the acceptance level and inspection method, and the measured distribution of fit-up variation from current production.
Then state what the supplier still has to prove: the pass plan derived from the groove rather than assumed, torch access at every pass position, whether seam finding or tracking is required given your fit-up data, and the dimensional tolerance class the finished assembly must meet.
Frequently asked questions
Why not weld a thick joint in one pass?
Because the heat input required would cost more in grain growth, heat-affected zone width and distortion than the additional passes cost. Filling in layers keeps each pass within a workable heat input range.
Who should decide the number of layers and passes?
It follows from the groove geometry, which sets the cross-section to be filled. It is a planning decision made before production, not a choice made at the torch.
Does robotic welding need seam tracking for multi-pass work?
It depends on your fit-up variation. A robot executing a taught path does not adapt to a variable bevel the way a welder does, so either the preparation is controlled tightly enough for the taught plan or sensing is added. The measured variation from current production decides which.
Does a good-looking weld mean the joint is sound?
No. Surface appearance shows bead sequence, overlap and regularity. Penetration, inter-run fusion and mechanical properties require the corresponding examination and procedure qualification.
Project inputs for an application review
Send the following and the pass plan can be reviewed against the real joint instead of a parameter set:
- plate thickness and material, with the joint types in scope
- the groove preparation you use now, or intend to use, and how it is produced
- the acceptance level and the inspection method that will demonstrate it
- measured fit-up and bevel variation from current production
- the dimensional tolerance class the finished assembly has to meet
If you are planning robotic multi-pass welding on medium-thick plate, send the thickness and material, the joint types and groove preparation, the acceptance level and inspection method, your measured fit-up variation, and the dimensional tolerance class required. The layer and bead strategy can then be derived from the groove rather than assumed. Related reading: multi-pass process planning and interpass control, structural welding applications, automatic welding system solutions, datum recovery on large structures.