Troubleshooting Yaskawa UP6 Welding Point Variation: Complete Field Diagnostic Guide

The Yaskawa UP6 (Motoman UP6) is a six-axis articulated arc-welding robot rated for a 6 kg payload. After 20,000–30,000 servo-on hours, the platform begins to exhibit weld-point variation that cannot be corrected by TCP (Tool Center Point) re-mastering alone. This reference walks through a structured diagnostic flow that separates mechanical, servo/feedback, process, and calibration root causes, and provides the verification checks a maintenance engineer should perform on a production line of multiple UP6 cells.

1. Problem Definition and Acceptance Limits

Before any diagnosis, quantify the variation. A spot-weld position deviation of even a few millimetres dramatically reduces weld nugget quality. Independent industrial studies have measured deviations of spot weld positions of up to 19 mm on production bodies, with most of that variation attributable to robot repeatability and locating-fixture tolerance rather than the weld timer itself.

Record the variation per axis. PAM (Position Adjustment Mode) shifts a taught point by a user-defined vector, but it cannot compensate for a varying error because the error itself is non-deterministic. The objective is to remove the variation, not mask it.

2. Quick Triage: Reference-Point Insertion Test

The fastest way to confirm whether the variation is real is to embed two non-welding reference moves at the start and end of every production program. These moves should target a fixed mechanical feature of the fixture (a dowel pin, a machined step, a clamped reference block) and pause for a defined dwell (e.g. 0.5 s) or set PL=0.

1. Open the program in the teach pendant editor.
2. Insert a move to the reference feature with no welding output (arc OFF, wire OFF).
3. Set a dwell or output DO[…]=ON for 0.5 s so the robot stabilises before the point is reached.
4. Add a duplicate move at program END pointing to the same feature.
5. Run 20–30 cycles and visually log whether the robot lands on the reference every time.

If the robot misses the reference, the variation is in the manipulator — not the weld process. If the robot lands consistently but the weld drifts, the problem is downstream (wire feed, contact tip, gas, fixture). This single test separates the two failure domains in under an hour.

3. Mechanical Inspection Procedure

At 27,000 servo-on hours the UP6 has typically reached the second or third grease interval and is past the first harmonic-drive inspection milestone. The following sequence is performed with the controller in Teach mode, servo OFF.

3.1 Anchor and Base Bolts

Vibration from spot-welding currents and fixture clamps slowly loosens the four M20 base anchor bolts. Use a torque wrench set to the OEM specification (typically 320 Nm for the UP6 base) and re-torque in a cross pattern. Check the grout pad for cracks radiating from the anchor.

3.2 Torch Holder, Anti-Collision Clamp, and Cable Dressing

The torch holder clamp and the anti-collision device (if fitted) are the most common source of single-axis drift on this generation. With the torch removed:

1. Loosen the anti-collision module and re-clamp it to the rated torque.
2. Inspect the dowel pin in the anti-collision body for shear or rotation.
3. Check the dress pack clamp at the B-axis and T-axis for slip.
4. Push-pull the welding cable and confirm zero axial play at the connector.

3.3 Gearbox Backlash Check (S, L, U, R, B, T)

With servo OFF, manually articulate the robot to a configuration where the L and U axes are approximately perpendicular to the floor and the B axis points down. Push the TCP laterally with moderate, controlled force — never impact load — and observe how the arm returns to the rest position.

The S, L, and U axes carry the highest static load and are the most likely first-failure points. R, B, and T are lighter but the T-axis (wrist roll) is exposed to welding-cable torsion and is a frequent second-stage failure.

3.4 Mechanical Fortification Pass

Walk the cell and torque every externally accessible bolt: anchor, base, J1 (S) housing, J2 (L) shoulder, J3 (U) shoulder, counterweight retention, transformer mount, and fixture clamps. Replace any bolt that shows necking or thread deformation. This is the single highest-yield 30-minute activity for stabilising a 27k-hour UP6 cell.

4. Servo Feedback and Pulse-Error Diagnostics

The Yaskawa NX100/DX100 controllers expose a per-axis pulse-error (position-following error) display. With servo ON and the robot fully stopped, navigate:

TOP MENU → ROBOT → MONITOR → SERVO ERROR (or PULSE COUNTER page)

4.1 Reading the Pulse Counter

A reading of ±3 or more indicates the closed loop is over-correcting. Motor current rises, the axis hunts, and position may be lost at the next stop. Root causes, in order of frequency on a 27k-hour UP6:

1. Encoder backup battery voltage low (NX100 = 3.6 V lithium; replace on a 3-year cycle).
2. Encoder cable shielding broken at the J3 or J4 bend.
3. Harmonic-drive backlash that the position loop is trying to absorb (see Section 3.3).
4. Servo-pack gain drifted — only re-tune after mechanical causes are eliminated.

4.2 Axis Isolation via Joint Jog

Use the Joint coordinate system, not World or Tool, to move one axis at a time. If a taught point drifts in Z as well as X or Y, the U-axis is suspect (it moves the TCP in an arc). The standard mapping is:

Jog the suspect axis +100 mm and −100 mm; if the reference point returns to the same XY after a return move, the axis is mechanically healthy. If the return is off, the variation lives on that axis.

5. Welding-Process Variables That Mimic Robot Variation

Two identical-looking weld deviations — one from the robot, one from the process — must be separated. Lincoln Electric's process reference lists the primary penetration variables:

• Travel speed — too slow widens the bead, too fast reduces penetration.
• Voltage — directly controls bead width and toe angle.
• Amperage — controls penetration depth and deposition rate.
• Electrode angle — push vs drag angle shifts bead shape.
• Contact-tip-to-work distance (stick-out) — increases electrical resistance, changes arc length, and — critically — allows the soft MIG wire to curl through 360° as the robot travels.

The single most overlooked variable on an aging UP6 cell is wire stick-out. A 0.8 – 1.2 mm MIG wire should be held at approximately 13 mm (1/2 in) from the contact tip to the workpiece. Longer stick-out makes the wire act as a curved spring; as the torch approaches the joint, the curl direction shifts and the apparent arc landing point rotates through the full 360°.

5.1 Contact Tip and Nozzle Condition

Replace the contact tip on a scheduled interval, not on failure. A spalled or eccentric tip can shift the electrical pick-up point by 1–2 mm. Verify the tip is concentric with the nozzle bore using a pin gauge or visual reference. A tip that is proud of the nozzle is easier to set but is also more vulnerable to crash damage — use a tip that is flush or 0.5 mm recessed for crash-tolerant cells.

5.2 Wire Feed and Liner

Check the liner for kinks at the feed-roll entry and at the torch neck. A kinked liner causes a saw-tooth feed pattern that the arc interprets as a moving contact point. With the wire disconnected, pull the wire through the liner by hand — it should slide with a consistent low drag. Replace the liner every 1,000–1,500 servo hours of arc time, sooner in dirty environments.

6. Maintenance Schedule Reset

A UP6 at 27,000 servo-on hours has likely missed one or more scheduled services. The standard Yaskawa service intervals for the UP6 / DX100 platform are:

Use only the grease specified in the Yaskawa maintenance manual (typically a polyurea or lithium-complex NLGI 2 for the S/L/U axes, and a different grade for the wrist axes). Mixing greases causes channeling and accelerated wear. The hour-meter reading is in TOP MENU → ROBOT → STATUS → HOUR METER.

7. Zeroing and Absolute Reset Procedure

The procedure is performed under Management mode:

1. Switch the mode selector to MANAGEMENT and enter the security code.
2. Navigate to TOP MENU → ROBOT → HOME POSITION (or ZEROING POSITION on later firmware).
3. For each axis, manually align the visible witness mark on the motor housing with the corresponding mark on the adjacent structure. The marks are scribed lines on the harmonic-drive housing.
4. Confirm axis-by-axis using the pendant; the controller will write the absolute reference on completion.
5. Re-master TCP and verify every critical program point with a dial indicator or a calibrated reference target.

If you cannot reconcile the variation before zeroing — for example, if the backlash test in Section 3.3 still shows a non-recoverable offset — the absolute reset will not fix the problem and may even amplify it. Always complete the mechanical and pulse-error checks first.

8. Verification Procedure

Once the corrective work is complete, run a structured acceptance test before returning the cell to production.

1. Static repeatability: Run the reference-point program from Section 2 fifty times. Record the TCP XY on a dial indicator or vision target. Acceptable is < ±0.3 mm on a healthy UP6.
2. Dynamic repeatability: Run a representative production program at full speed and 70% speed; record the deviation of the same reference at the end of program.
3. Pulse counter steady-state: With the robot stopped, all axes should report −1, 0, or +1. Anything outside this band is a reject.
4. Cross-load test: Run the program on the heaviest part in the production mix; check that pulse error does not exceed +2/−2 under load.
5. Process cross-check: Cut five coupons from a representative part and section the welds. Compare bead profile to the original WPS (Welding Procedure Specification). If the bead profile is correct and the robot is repeatable, the line is healthy.

9. Diagnostic Decision Matrix

10. Common Pitfalls on a Six-Robot UP6 Line

With six identical robots on one line, two maintenance disciplines save significant time:

1. Per-robot hour meters. Don't assume the line is synchronised. Robots 1 and 6 may differ by 5,000+ hours of arc time. The robot with the most arc time is statistically the first to fail.
2. Shared fixtures, individual dress packs. A new dress pack on a worn liner is wasted money. Replace liner, contact tip, and gas diffuser as a set, and log the date on the cable label.

PAM is a legitimate production tool for compensating known offsets (e.g. a family-of-parts where each part has a measured shift). It is not a substitute for a healthy manipulator. If the underlying variation is wider than the PAM shift you would need to apply, stop using PAM on that point until the root cause is removed.