The Rail Axis: Verify the Boundaries Before Quoting Coverage
Direct answer: A robot rail external axis extends the working envelope along the travel direction and leaves every other direction unchanged. Each change of station adds a locating operation and another error source, and the junction between stations is where a long seam first shows a step. Covered length and weldable length are therefore different questions, and the boundary has to be verified before either one is quoted to a customer.
Who this is for: Engineers specifying ground rails and external axes for welding cells handling long structural members.
Scope: This EVST guide is written from one filmed station: a long structural member fixed on a welding table, with a robot travelling on a ground rail and welding a length of seam at successive positions along that travel. Three takes of that one station are used. Nothing here states a cycle time, a repeatability figure or an acceptance result for the filmed work.
One direction gets longer, the others do not
A ground rail extends the robot’s working envelope along the axis of the rail. It does not extend the others, and reading it as a longer robot is what produces reach errors perpendicular to the track.
The usable volume is closer to a prism swept along the rail: the cross-section is still the original working range. Any part of a workpiece outside that cross-section stays out of reach no matter how far the carriage travels.
That distinction matters at proposal stage, because a member that is long but also deep or tall can sit inside the rail’s coverage and still fall outside the cross-section at the positions where its seams are. Checking the envelope against the swept cross-section, rather than against rail length, is what catches it.
According to ISO 9283:1998, pose accuracy and repeatability are defined for stated load and velocity conditions, which is why a robot rail external axis is assessed under the load it will actually carry.
The part stays, the robot moves
In the reference footage the structural member is fixed on a welding table and the robot travels along the rail beside it. That division of labour suits parts that are long, heavy and awkward to turn over.
Where the part is easy to rotate, the opposite division is often better value: let the part move and keep the robot fixed. Which of the two applies is one of the first things to establish, because it decides the shape of the whole cell.
Weight is usually the deciding factor rather than length. A member that is long but light can often be carried on a positioner; one that is long and heavy is easier to leave on the table and travel past.
According to ISO 13920:1996, tolerances are stated against reference features, and in practice each station of a robot rail external axis re-establishes those features for the length it covers.
Robot rail external axis: every station is a new location
The robot stops at successive positions along the travel and welds a length of seam at each. Every change of station is a fresh locating operation, and every locating operation is another source of error.
Errors accumulate with station count, which makes the number of stations a parameter to balance rather than to maximise. Fewer stations means each must cover more, which raises the demand on posture; more stations means more locating operations. Both directions cost something, and the balance is calculated for the particular member.
The balance also shows up in cycle time. Each station change costs travel and re-location, so a layout that minimises station count for accuracy reasons often improves cycle at the same time; where the two pull apart, accuracy governs, because a step in the seam cannot be recovered later.
The junction between two stations
The short length of seam between two stations is where the result becomes visible first. If the two do not meet, the long seam carries a step in the middle of it.
There are two usual approaches. Overlap a short length between stations, or find a natural break in the structure and divide the seam there.
Overlapping requires the locating consistency between stations to be good enough, otherwise the overlap region shows double deposition or misalignment. A structural break is less demanding on accuracy but needs the structure to allow it, for example at a stiffener or a gusset where a division is reasonable both visually and structurally.
Which one applies follows from the measured repeatability and the required weld quality level. It is a quantified decision rather than a habit, and it belongs in the proposal, because it drives station count, which in turn drives cycle and rail length.
According to ISO 5817:2023, the junction between two welded lengths is assessed by the same imperfection limits as the rest of the seam, so a robot rail external axis is only as good as the way its stations meet.
A decision table for a rail axis
Each row is settled from the customer’s own member and tolerance. Rail length alone answers none of them.
| Decision | Settled from | What goes wrong if skipped |
|---|---|---|
| Which direction is extended | Workpiece envelope against the swept cross-section | Reach miscalculated perpendicular to the rail |
| Part moves or robot moves | Member length, weight and whether it can be turned | A rail bought to solve a positioner problem |
| Station count | Seam layout, accumulated locating error | Error accumulates or posture demand rises |
| Junction method | Measured repeatability, weld quality level | A step left in the middle of the long seam |
| Repeatability verification | Acceptance method, loaded and unloaded | Coverage quoted as if it were weldable length |
What the footage cannot show
Repeatability is not visible in a moving picture. ISO 9283 defines how pose repeatability is measured, and it has to be measured loaded: unloaded figures look better than the condition in which the robot works, carrying a torch and cable pack at the end of the travel.
Rail straightness and the foundation matter for the same reason. Small undulations in the track mean the same program sits at slightly different heights at different stations.
Cycle time, station changeover time and inspection conclusions for the whole seam are equally outside what footage carries.
The segments used here are three takes of one station, not three lines.
The order worked through is the same each time: establish which direction actually needs extending, decide whether the part or the robot moves, then verify the boundary before coverage is quoted.
Frequently asked questions
Does a longer rail mean a longer weldable seam?
Not by itself. Coverage is a travel question; weldable length depends on locating consistency between stations and on how the junctions are made.
How many stations should a long member use?
As few as posture allows. Each additional station adds a locating operation and an error source, so the count is balanced against how much each station can cover.
Should repeatability be measured loaded or unloaded?
Loaded, and at the end of the travel. That is the condition the robot actually works in, and it is where the difference from the unloaded figure is largest.
When is a positioner the better answer than a rail?
When the part can be turned. Rotating a member is usually cheaper than moving the robot along it, and it avoids adding a locating operation per station.
Project inputs for an application review
To assess a rail-based cell, an EVST application review opens from the following inputs:
- member length, section and weight per part
- seam positions, and whether seams cross more than one station
- required repeatability and the acceptance method
- existing rail travel and carriage arrangement, and whether the part can be rotated
Send the member length, the weight and the seam positions, and we will work through travel, station division and junction method against your own accuracy requirement. Related reading: welding robot workstation presentation, EVST robot product overview.