Robotic Spot Welding: Count the Welds Before the Robot
Direct answer: A robotic spot welding plan starts at the part, not at the robot. Count the welds on one part, map where they sit, and decide whether they must stay separate or can be merged into longer runs. Weld count and spacing govern the cycle far more than peak arm speed, and one shared fixture datum is what makes each weld position repeatable.
Who this is for: This guide is written for manufacturing engineers and buyers evaluating automation for sheet-metal frames and similar fabricated parts, where one part carries many short welds rather than a few long seams.
Scope: It covers how weld count, weld distribution, the fixture datum and torch access decide whether the cell is worth building, and what each judgement still owes in evidence. It does not cover resistance welding gun sizing, weld metallurgy, or the finishing operations that follow.
What a robotic spot welding plan starts from
A note on the term first, because it is used loosely. In this article robotic spot welding means a robot placing many short, discrete welds on one part, which is the pattern in the reference footage: a thin-sheet frame built from an array of flat bars, with the arm starting and stopping the arc at each position in turn. It is not resistance welding with a weld gun, and the planning questions are different.
The usual first question in an enquiry is how fast the arm needs to be. That question is almost always premature. What decides whether the cell is worth building is the part: how many welds it carries, where they sit, how the part is held, and which poses the torch can actually use to get to each one.
According to ISO 10218-2:2025, it is the robot application rather than the robot on its own that has to be assessed, and a welding cell behaves the same way commercially. The arm is one component in a selection that also contains the fixture, the torch, the power source and the loading method, and the part is what sets the requirement for all of them.
In practice the cheapest hour on a project like this is the one spent listing the welds before anyone opens a robot catalogue. A weld list with positions, sheet thickness and material is enough to tell whether the application is comfortable, marginal or unsuitable, and that judgement is far cheaper on paper than on the floor.
Weld count and spacing set the cycle
The time to weld one part is made of three things: the arc time itself, the travel between weld positions, and the starts and stops at each one. When a part carries many short welds, the first of those is often the smallest term. The arm spends most of the cycle moving and settling rather than welding.
That is why weld distribution matters more than the headline speed of the arm. Welds spread across a large frame produce long travel moves and many approach and retract motions; welds grouped along one edge produce short hops. Two parts with the same weld count can therefore behave very differently, and no amount of arm speed closes that gap.
The starts and stops carry their own cost, and it is not only time. Every arc start is a moment where the process has to become stable quickly, and a part with many short welds asks for that repeatedly. Where the welds are short, start behaviour deserves more attention during trials than peak travel speed does.
This leads to the one question worth asking before any equipment discussion: do these welds have to stay separate, or can some of them be merged into longer runs. The answer changes the cycle, the distortion behaviour and sometimes the whole cell concept, and it belongs to the part designer as much as to the automation engineer.
The fixture datum decides repeatability
In the reference footage the flat bars are laid out on the table to a regular pattern. That regularity is doing more work than it appears to: it is what allows the same program to land on the same weld position on the next part. A robot repeats its own motion very consistently, so if the part arrives in a slightly different place, what the cell repeats consistently is the wrong position.
On thin sheet the fixture has a second job beyond location. Assembly gaps, sheet flatness and where the part is held down all affect how a short weld forms, and none of those are things a program can correct after the fact. Where a weld looks inconsistent from part to part, the fixture and the incoming part are the first two places to look, not the parameters.
According to ISO 3834-1:2021, the level of quality requirements applied to a fabrication is itself a selection, driven by the criticality of the product rather than by the equipment used to make it. That framing is useful here because it separates two decisions that often get merged: what this joint has to achieve, and what the cell needs in order to achieve it repeatably.
A practical check needs no instrumentation. Load and clamp the same part several times, and look at how far the weld positions move between loads. If that scatter is stable and small relative to the joint, the concept is worth costing. If it is not, the fixture is the project, and EVST would rather resolve that before an arm is fixed than after.
Decision table: what the weld map justifies
The table maps what a weld list and part family can justify on their own against the evidence a supplier still has to produce.
| Evidence from the application | Selection it justifies | Evidence you still owe |
|---|---|---|
| One part family, welds grouped in a small area, stable incoming parts | A single-station cell sized on travel between welds rather than on peak arm speed | Trial parts welded from the real fixture, with weld positions checked across repeated loads |
| Many welds spread across a large frame | A layout study before arm selection, since travel and approach dominate the cycle | Pose check at the least accessible weld, with the real torch and cable package fitted |
| Several part variants sharing one table | A fixture concept with a common datum and changeable locators, costed as part of the cell | Changeover trial on the real variants, and a weld position check after each change |
| Weld list still open, or joint design under discussion | No arm selection yet; a shortlist with the sensitivity of each option recorded | A frozen weld list with positions, thickness and material before any commitment |
Torch access and the poses you can really use
Reach on a datasheet is a radius measured in free space. The reach a welding cell needs is a set of specific torch poses at specific positions on a part that is clamped inside a fixture. Between the two sit the torch body, the cable package, the clamps and the table itself.
The demanding position is rarely the furthest one. It is usually the weld that sits closest to a clamp, in a corner, or under an overhanging feature, because there the torch angle is constrained from several directions at once. Checking the easy positions and assuming the rest is how a cell arrives on site and then needs its fixture reworked.
This is also where the tooling concept and the robot welding workstation configuration have to be discussed together rather than in sequence. A slimmer torch may open access that a larger arm cannot; a different clamp position may do more than either. Those trades are only visible when the fixture and the arm are on the same drawing.
What the footage proves and what it does not
The reference footage proves a visible process. A six-axis arm carrying a gas-shielded torch completes multiple short welds in sequence on one thin-sheet frame; the bars are laid to a regular pattern on a fixture table; the arc starts and stops cleanly at each position. Those are observable facts about the operation.
It proves nothing about cycle time, weld strength, distortion or first-pass yield, and no honest reading of it can supply those. According to ISO 5817:2023, quality levels for imperfections in fusion-welded joints are defined for the joint and assessed on it, which is precisely why they cannot be inferred from watching a weld being made.
Keeping visible evidence and derived conclusions apart is the fastest way through a selection, because it makes the remaining questions explicit instead of leaving them to be discovered during acceptance. EVST keeps that separation in writing so a comparison between suppliers stays a comparison of the same thing.
What to put in the enquiry
State the part: a drawing, the weld list with positions, sheet thickness and material, and how many variants share the station. Add the batch pattern, because a part family that changes several times a shift is a different cell from one that runs all week.
Then state what the supplier still has to prove: weld position scatter across repeated loads on the real fixture, a pose check at the least accessible weld with the real torch fitted, and a changeover trial if variants share the table. Add the surrounding constraints as well, since loading method, operator access and the way this station sits inside a wider robotic factory integration all shape the answer. A request written this way is comparable across suppliers; a request that asks only for a robot product range and a price is not.
Frequently asked questions
Is this the same as resistance spot welding?
No. Here the robot places many short arc welds in sequence on one part, which is what the reference footage shows. Resistance welding with a weld gun is a different process with different tooling, access and power requirements, and its planning questions do not transfer.
Can we compare robots on speed alone?
Not usefully. Where a part carries many short welds, the cycle is dominated by travel between positions and by arc starts and stops. Two arms with the same quoted speed can produce different cycles on the same part depending on how the path is planned and how quickly the process stabilises.
How do we know the fixture is good enough?
Load and clamp the same part repeatedly and look at how far the weld positions move between loads. Stable, small scatter relative to the joint means the concept can be costed. Large or unstable scatter means the fixture is the project, and no program change will substitute for it.
Does the footage show the cycle time we can expect?
No. The footage is evidence of a visible process: a robot completing multiple short welds on a clamped sheet frame. Cycle time, distortion and acceptance results have to come from trials on your own parts, and any figure quoted without those trials is an estimate rather than a measurement.
Project inputs for an application review
Send the following and the concept can be reviewed against the real part instead of against a headline speed figure:
- the part drawing and the weld list, with position, sheet thickness and material
- how many variants share the station, and how often they change
- your fixture concept, including how the part is located and held down
- the loading method and the operator access the station has to keep
- the cycle requirement, so travel and arc starts can be checked rather than assumed
If you are evaluating automation for a part that carries many short welds, send the part drawing and weld list, sheet thickness and material, the variant and batch pattern, your fixture concept, and the loading and access constraints around the station. The concept can then be checked at the least accessible weld rather than at an average one. Related reading: robot product range.