The robot has hit the tooling, a guard, a misplaced part, or even an operator. The line stopped on emergency, and now there's pressure to reset the alarm and get back to production. That's exactly the moment when most chronic repeatability faults get created — the ones that show up weeks later with no obvious cause. A collision isn't just a scare: it can leave a gearbox with play, a misaligned encoder, or an axis cable with damaged insulation, and none of that is visible to the naked eye.
Why resetting the alarm and moving on isn't enough
The controller only detects what its sensors measure: motor overcurrent, position loss, or a limit switch trip. It won't detect an incipient backlash in a gearbox, a bearing that took a hit, or a TCP shifted by half a millimeter. Those problems surface later as repeatability loss, vibration in specific trajectory segments, or parts falling out of tolerance — and by then it's hard to link the fault back to the original impact.
Most common causes of collisions on the floor
- Programming error: a badly taught point, wrong offset, or tooling change without updating the program.
- Part or pallet out of position due to a sensor or fixture failure.
- Excessive speed in automatic mode in a tight-clearance zone.
- Manual intervention in teach mode without reducing speed.
- Failure of a safety guard or interlock that didn't stop the robot in time.
Inspection checklist before restarting production
1. Safety and visual check first
Before touching anything, check for oil leaks at the gearboxes, loose or pinched cables, displaced wrist guards, and the condition of the end effector (gripper, welding gun, tool). If there was contact with a person, follow the plant's incident protocol before any technical check.
2. Mechanics: play, alignment, fasteners
Move each axis manually at low speed and look for abnormal play, new noise, or irregular resistance. Check the tightness of the base, wrist, and end-effector fasteners; a collision usually loosens bolts before it breaks internal parts.
3. Brakes, motors and noise
Listen to each axis while moving slowly: a screech, a knock, or braking that feels different than usual points to possible damage in the brake, motor, or gearbox. Compare motor temperature after a few minutes of movement with the cell's normal behavior.
4. Repeatability and mastering
Send the robot to a known reference position (home or a saved teach point) several times and compare the deviation. If the robot doesn't repeat within factory tolerance, suspect a gearbox impact or loss of mastering — not just a program adjustment. This is where the most time gets wasted if teams try to "fix it with offsets" instead of checking the mechanical cause.
5. TCP and tool offsets
Verify the TCP using the brand's standard method (fixed point, four-point, etc.). A TCP shift after an impact on the end effector is one of the most common faults and one of the biggest sources of quality rejects when it goes unnoticed.
6. Cables and connectors
Check axis cables and their connectors in any section that may have taken the impact or a sharp pull. A cable with damaged shielding may not fail immediately but can cause intermittent interference weeks later.
7. Dry run before real production
Run the full program in manual mode at reduced speed, without a part, before switching to automatic. Only once the full cycle repeats cleanly should you return to normal production.
Warning signs you shouldn't ignore
- New vibration in a specific trajectory segment that wasn't there before.
- Position or axis-tracking alarms appearing intermittently after the collision.
- Parts starting to fall out of tolerance in one area of the cell, even though the program hasn't changed.
- An oil leak at a gearbox that wasn't present before the incident.
Common mistakes after a collision
The most common one is resetting the alarm and going back to production with no mechanical check, trusting that "if the robot moves, it's fine." The second is failing to document the incident and the inspection performed, which makes it impossible to link future failures back to this event. The third is trying to correct repeatability loss by adjusting the program instead of checking for physical damage in the gearbox, encoder, or TCP.
When to stop and call a specialist
If you detect mechanical play, new noise, out-of-tolerance repeatability, or a suspected gearbox impact, stop the cell and call a technician. Continuing to run with a damaged gearbox accelerates its total failure and can end in a far more expensive emergency replacement than a scheduled repair; see the full process in our article on when and how to replace a gearbox. To check whether the issue is calibration-related, it's also worth reviewing mastering and axis calibration. If the incident was serious or recurring, a technical audit helps identify whether the root cause is mechanical, programming, or cell safety related, and our corrective maintenance service can step in with part replacement or exchange if the damage requires it.