Commissioning, Maintenance and Troubleshooting
Mastering and calibration, acceptance tests, backups, encoder batteries, lubrication, spares, payback and a troubleshooting table.
Phases of a project
| Phase | What happens | Output |
|---|---|---|
| Concept and feasibility | Define the task, the rates, the part variation; reach study; rough cost and payback | Requirements and a go / no-go |
| Design | Choose robot and tool, layout, safety concept, PLC and I/O design, simulation of the cycle | Drawings, specification, risk assessment |
| Build and factory acceptance test (FAT) | Build the cell at the integrator, test the tool, the program and the safety with real parts | FAT report, punch list |
| Install and commission | Set in place, connect services, calibrate, teach, test, tune | Cell running at reduced and then full rate |
| Site acceptance test (SAT) and handover | Run to the acceptance criteria at the customer, train staff, hand over documents | SAT report, acceptance certificate |
| Operate and maintain | Production, maintenance, monitoring, continuous improvement | A reliable, documented cell |
Commissioning step by step
| Step | What to do | Check |
|---|---|---|
| 1. Mechanical and electrical checks | Base bolted to spec, tool fitted, cables routed with a service loop, earthing, air pressure, network, safety circuit wired | Torque of the bolts; insulation and earth continuity; the emergency stops work with the robot off |
| 2. Power up and master | Power on, check for alarms, confirm the axes are mastered (zero position) and the encoder data are valid | Move each axis slowly through its range, in jog; check the limits |
| 3. Tool, payload and frames | Enter the tool center point, the load data, the frames; measure the TCP (see Coordinate Frames and Kinematics) | Reorient the tool about the TCP: the tip stays put |
| 4. Brakes and stopping | Brake test; measure the stopping time and distance at the worst conditions (see Safety: Standards, Safeguarding and Cobots) | Results are within the values used in the safety distances |
| 5. Safety functions | Test every gate, curtain, scanner and emergency stop; test the reduced speed and the enabling switch | Each device stops the robot as the risk assessment requires; the record is signed |
| 6. I/O and handshake | Force each signal and check it at the robot and the PLC; run the handshake in manual (see Cell Design, PLC Integration and Feeding) | Every signal on the list is right; the fault signals work |
| 7. Teach and test the program | Teach points in frames, run slowly in test mode, step through, then at 10, 25, 50, 100 % speed | No collisions; the positions repeat; the cycle time is as expected |
| 8. Vision and sensors | Calibrate the camera and hand-eye; check the grip sensors (see Vision, Calibration and Accuracy) | Verify by touching known points |
| 9. Optimize | Blend, tune acceleration, overlap gripper actions (see Payload, Inertia, Speed and Cycle Time) | Cycle time reached without overload warnings |
| 10. Run-in, back up and hand over | Run production, fix what shows up, back up everything, train the operators | Acceptance criteria met |
Worked example: why mastering matters
If an axis is mastered 0.10° out, the tool at 750 mm reach is off by 750 × 0.001745 = 1.31 mm, and every taught point made before the error is now off by about that much. Master after any motor, gearbox or encoder work, and after a flat encoder battery (see below); re-check at least a few key points afterwards.
Worked example: the brake test
A joint carries a holding torque of 80 N·m (arm and payload) and its brake is rated 120 N·m: a margin of 1.5. The brake test holds the axis against a drive torque slightly above the holding torque with the brake applied; if the axis creeps, the brake is worn, and the robot must stop and not be used until it is changed. Test at the interval the manufacturer states, and always after any brake work.
Acceptance and availability
Agree the acceptance criteria before the build: the cycle time, the rate, the quality (reject rate, accuracy), the availability, and the safety validation. The usual measure of reliability is availability A = MTBF ÷ (MTBF + MTTR), where MTBF is the mean time between failures and MTTR the mean time to repair.
Worked example: availability
- A cell fails on average every 1 200 hours (5.0 times in 6 000 hours a year) and takes 3 hours to repair: A = 1 200 ÷ (1 200 + 3) = 99.75 %, 15 hours of downtime a year.
- With an 8 hour repair (no spare, a long wait for a technician) the availability falls to 99.34 %, 40 hours a year. Repair time matters as much as failure rate: spares, diagnostics and a trained person on site cut the downtime.
- A target of 98 % allows 120 hours of downtime a year: both cases meet it, but the second only just.
- A demonstration run. To show, with 95 % confidence, that a cell fails on fewer than 0.1 % of cycles, it must run 2 994 cycles with no failures (n = ln(1 − 0.95) ÷ ln(1 − 0.001)): at a 2.4 s cycle that is 2.0 hours. For 1 % it is only 298 cycles. Write the run length into the contract and agree what counts as a failure.
- Define each measure: availability, the OEE (overall equipment effectiveness: availability × performance × quality), what is excluded (planned maintenance, no parts supplied).
- Log the events from day one: every stop with a time, a reason and a duration (see Data Logging); the pattern shows where to improve.
- Agree the punch list (open items) at SAT, with dates and owners.
Backups and change control
| Back up | Why |
|---|---|
| Robot programs and system data (full controller backup) | To restore after a fault or controller replacement |
| Tool, load and frame data; calibration and mastering data | Positions mean nothing without them |
| I/O configuration and fieldbus settings | The handshake depends on it |
| PLC program and HMI | The other half of the cell |
| Safety configuration with its checksum | To prove that the validated settings are in place |
| Vision applications and calibration files | Cameras and models |
| Documents | Drawings, risk assessment, FAT and SAT reports |
- Back up before every change and after every good state, with a date and a note; keep copies off the controller (a PC, a server) and test a restore at least once.
- Control who can change a program: passwords and user levels on the controller, and a log of changes.
- A safety setting change needs the safety validation repeated (see Safety: Standards, Safeguarding and Cobots).
- Label the backup media with the cell and the controller’s software version; the restore needs the same version.
Preventive maintenance
| Item | Typical interval (check the manual) | What to do |
|---|---|---|
| Daily or each shift | Each shift | Look for leaks, noise, loose cables; check the tool and the grippers; check the safety devices’ indicators; clear dirt from the gripper and vacuum filters |
| Weekly | Weekly | Check hoses and cables on the arm; test the emergency stops and the door interlocks; clean the vision lenses |
| Grease or oil in the gearboxes | About 20 000 hours (3.3 years at 6 000 hours a year) or the maker’s time | Use the exact grease or oil; replace by the manual’s procedure; dispose of the old safely |
| Encoder backup battery | About every 3 years, or as the controller warns | Replace with the controller powered, so that positions are not lost (see below) |
| Cables and dress pack | By flex cycles (see below) | Replace the moving cables before they fail |
| Brakes | Brake test at the stated interval | Test; replace a worn brake |
| Controller and cabinet | Yearly | Clean filters and fans, check connections and the cabinet’s temperature, back up |
| Safety system | At least yearly and after any change | Test every function; record the results |
| Tool wear parts | By count or wear | Vacuum cups, finger pads, welding tips: change on a schedule |
Worked example: the dress pack
A cable on the wrist that flexes once for every four robot cycles and has a rated life of 5 000 000 flex cycles, in a cell that runs 6 000 hours a year at a 2.4 s cycle and 80 % utilization, flexes 1.8 million times a year: the life is 2.8 years. Replace the cable at about 70 % of that, 1.9 years, in a planned stop. At $1 800 that is a cost of about $648 a year, against the cost of a stop in the middle of production and the risk of a fault that looks like something else. Cables and hoses on the arm are the most common wear-out failure.
Encoder batteries and losing position
Many robots store the absolute position of each joint in the encoder with a small backup battery. If the battery runs flat or the encoder is disconnected with the controller off, the position is lost and the robot must be re-mastered (re-zeroed), which means re-checking all the taught positions. A battery costs about $40; the loss of the position costs hours. Replace on schedule, and do it with the controller on.
Condition monitoring
The robot controller already measures the motor current, speed, temperature and position error of every joint. Trends in these figures warn of a wearing gearbox, a failing brake, a tight cable or a heavier tool before a failure. Log them on a regular test move that always runs the same path with the same load (see Data Logging).
Worked example: a rising motor current
- The mean current of one axis on the standard test move, weekly: 3.10, 3.12, 3.09, 3.14, 3.18, 3.25, 3.33, 3.44, 3.58, 3.76 A. For the first five weeks it averages 3.13 A.
- From week 6 it rises, and by week 10 it is 20 % above the baseline. Set an alert at +25 % (3.91 A) and plan the inspection (the gearbox, the cable, the lubrication) before it is reached: a straight line through the 10 weeks, with a slope of 0.070 A per week, would cross it after about 14.2 weeks, although wear usually accelerates, so use it as a rough upper bound on the time you have.
- Use features that compare like with like: the same move, the same load and temperature. The methods for finding trends and anomalies are in Machine Learning (see its page on features for signals and machines).
Troubleshooting
| Symptom | Likely cause | Try |
|---|---|---|
| Robot misses the pick point | Wrong TCP or frame, loose fixture, part position varies, vision calibration drifted, a lost mastering | Re-measure the TCP and frame; check the fixture; recalibrate the vision; check the zero marks |
| Linear move stops or slows near a pose | Singularity or joint limit (see Coordinate Frames and Kinematics) | Change the approach pose, use a joint move, or re-teach with a different configuration |
| “Position unreachable” error | Outside the envelope, tool too long, wrong configuration | Check the distance to the base; try the other elbow or wrist configuration; move the part |
| Robot loses its position after power-off | Encoder battery flat | Replace the battery and re-master the robot as the manual directs |
| Part drops from the gripper | Vacuum or air supply low, leak, worn cups, speed too high (see End-of-Arm Tooling and Grippers) | Check pressure and vacuum at the tool; clean or replace cups; reduce acceleration; add a grip check |
| Overload or collision fault | Payload data wrong, tool heavier than set, a crash, a tight cable | Enter the correct tool mass and center of gravity; check for a collision and for a binding hose; reduce the speed |
| Vibration or noise from one axis | A worn gearbox or bearing, a loose tool, too high a gain or acceleration | Listen and log the current; check the tool mount; lower the acceleration; call the maker |
| Cycle slower than planned | Waiting for the PLC or a part, conservative speed settings, stopping at every point | Log where time is spent; use blending; check handshake delays (see Cell Design, PLC Integration and Feeding) |
| Safety circuit trips unexpectedly | Gate switch misaligned, curtain dirty or misaligned, a wiring fault | Check the safety relay diagnostics and the device LEDs; clean and realign |
| Communication fault with the PLC | Cable, switch, address clash, a changed setting | Check the link LEDs, addresses and the fieldbus diagnostics (see Industrial Communication) |
| Positions drift during the shift | Thermal growth, a loose fixture, a moving conveyor frame | Warm up the robot; check the fixture; re-measure the frames |
| Vision finds the part but the robot misses | Hand-eye calibration, wrong height, lens or camera moved | Re-calibrate; check by touching a known point (see Vision, Calibration and Accuracy) |
- Work from the alarm code and the log first; controllers record the last alarms with the axis and the time.
- Change one thing at a time and note it. Do not reset a fault repeatedly without finding the cause.
- Before entering the cell, stop the robot, lock out and take the key (see Safety: Standards, Safeguarding and Cobots); the arm may drop when the brakes are released.
- Keep spares for the failures that stop production: a teach pendant, a set of cables, grippers, vacuum cups, filters, a controller fan and an encoder battery.
Payback and the business case
A robot is an investment. The simple measure is the payback time: the installed cost divided by the net annual saving. A fuller view discounts the future savings (net present value).
Worked example: a palletizing cell
- Cost: $76 000 installed (arm, tool, fixtures, guarding, integration; see Robot Types, Specifications and Selection).
- Savings: it replaces the equivalent of 1.0 full-time operator across the shifts, at $48 000 a year each (loaded): $48 000; plus $6 000 of reduced damage and injuries: $54 000.
- Running costs: maintenance, spares and support at 6 % of the cost, $4 560, and energy, $100: a net saving of $49 340 a year.
- Payback = $76 000 ÷ $49 340 = 1.5 years (18 months). With an 8 % discount rate over 7 years the net present value is $180 882 (positive: worth doing).
- Check the soft factors: flexibility, quality and consistency, safer work, the ability to keep working when people are scarce, and the risk of the project. Include the costs of downtime and training, and of the people’s time during the project.
Moving, upgrading and retiring a cell
- Moving a cell: re-validate the safety; re-calibrate the frames and the vision; check the services and the floor at the new place (see Cell Design, PLC Integration and Feeding).
- Upgrades (a new tool, a faster rate, a new product) change the risk assessment, the payload data, the stopping times and the cycle: review each.
- Controller and software updates can change behavior: test on a spare or in simulation, and keep the old version’s backup.
- Retiring: remove stored energy, drain the gearbox oil, dispose of batteries, electronics and the oil by the rules, and sell or recycle the arm: robots keep value for a long time.
- Train a second person to run, back up and fix the cell, so that the knowledge does not walk out of the door with one individual.
Related
- Industrial Robotics: the overview of this guide.
- Safety: Standards, Safeguarding and Cobots: the validation tests at commissioning.
- Cell Design, PLC Integration and Feeding: the documentation and the cell design.
- Payload, Inertia, Speed and Cycle Time: payload data and cycle time.
- System Integration and Data Logging: commissioning the whole system and logging its events.
FAQ
What is mastering, and when do I need to do it?
Mastering (zero-position calibration) stores the encoder zero of each axis. Do it after a motor, gearbox or encoder is changed, after a flat encoder battery lost the position, and after a major crash. Check taught positions afterwards.
How often should I back up the robot?
After every change and every good state, and at least on a regular schedule. Keep copies off the controller and test a restore.
What kills a robot cell’s availability?
Not the arm, usually: the tool (grippers and vacuum), the cables and hoses on the arm, sensors, and the interface with the PLC and the line. Long repair times from missing spares or diagnostics make it worse.
How do I decide whether a robot pays back?
Divide the installed cost of the whole cell by the net annual saving (labor, quality, scrap, less running costs), and check the discounted value over the cell’s life. Include the soft factors, and the cost of the project risk.
References
- Robot manufacturers’ maintenance manuals and the controller manuals (intervals, mastering, brake and battery procedures).
- ISO 10218-2 (validation and information for use); ISO 55000 (asset management); IEC 61508 and ISO 13849-1 (safety-system testing).
- Nakajima, S., Introduction to TPM (OEE); Smith, D. J., Reliability, Maintainability and Risk, Butterworth-Heinemann (availability and MTBF).
Last reviewed: October 2026. Intervals, costs and reliability figures are typical and illustrative: follow the manufacturers’ schedules and use your own data.