TCP Drift After a Torch Collision
After a minor collision at the torch, a 6-axis robot with a 280 mm torch length produces welds that are 1.8 mm out of position along the seam normal. The maintenance team re-mastered axis 6 using the fixture marks, replaced the torch mount and re-ran the production program without touching the tool frame. The torch is mounted on a 15 degree adapter, and the user frame was taught before the collision. A pointed tip touched to a datum plate shows a repeatable 1.8 mm error in one direction. The robot payload is 12 kg with a 1.5 kg torch, and the controller payload data still lists the original 1.2 kg torch.
1. Distinguish the three candidate error sources (axis-6 mastering, tool frame, user frame) and state which one the evidence supports.
Read the worked answer
Axis-6 mastering error rotates the tool about axis 6. A pointed tip would trace a circle or an arc, not a single repeatable direction. The observed error is repeatable in one direction, which argues against a mastering error as the primary cause.
Tool-frame error produces a translation when the tool center point is misdefined, and because the mount and 15 degree adapter were replaced without redefining the frame, the TCP is still where the old torch put it. A constant offset in one direction, confirmed by touching the datum plate, is the signature of a tool-frame error.
User-frame error would shift all taught positions relative to the part, including the datum touch. It can coexist here because the user frame predates the collision, but it would normally move the seam path and the datum point together, which does not match a repeatable one-direction touch error.
The evidence best supports a tool-frame error introduced by the mount change. Re-mastering axis 6 may also have added a small error, but it is not the main effect.
2. Calculate the angular error that would be needed if the fault were purely rotational, and use the result to support or reject the mastering explanation.
Read the worked answer
The angular equivalent of a 1.8 mm displacement at a 280 mm radius is:
angle = atan(1.8 / 280) = atan(0.00643) = 0.368 degrees.
If axis 6 were mastered using marks on a circle of radius about 300 mm, an offset of roughly 1.9 mm at those marks would be needed to create that angle. Marks that far off would normally be visible to the person doing the mastering, and a rotational error would produce a direction change as the tool rotates during the weld, which is not described.
So the calculation does not rule out a small mastering error, but it makes a pure rotational explanation less likely than a pure translation. The pointed-tip result being repeatable in one direction is the stronger evidence: it says the TCP is displaced, not merely rotated. Correct the tool frame first, then re-check mastering.
3. State the correction sequence and the verification tests, including the payload issue.
Read the worked answer
Correction sequence
- Redefine the TCP with a 4-point or 6-point method, including the 15 degree adapter, and store it in the active tool frame.
- Update the payload in the controller to the actual torch mass plus cable and any collision sensor, roughly 1.5 kg or more, rather than the old 1.2 kg. A 0.3 kg error at 280 mm is small but affects acceleration and can degrade path accuracy at speed.
- Re-check the user frame with a 3-point method against the fixture datums, because it predates the collision.
- Re-master axis 6 only if the marks are still suspect, and record the result.
Verification
- Touch the pointed tip to the datum plate from four orientations and confirm the spread is small.
- Scribe a dry-run path and measure it against the joint centerline at three points.
- Run a first-off weld, measure the seam position, and compare with the pre-collision record.
- Log the payload and frame values so the next mount change triggers a redefinition.