A new robot on the shop floor is exciting right up to the moment it does something nobody planned. The cheapest time to find out what it does wrong is before production parts, people or expensive tooling are anywhere near it. Here are eight dry-run steps for commissioning a robot cell, ordered from least to most physical. They apply to arms, mobile units and mixed lines alike, and each one is about finding a surprise while it is still cheap.
1. Run the program in simulation first
Before anything moves, load the cell model and the program into the simulator and watch the whole cycle. Look for reach problems, awkward wrist angles, collisions with fixtures and cycle times that look optimistic. A simulation does not replace the real cell, but it catches the errors that would otherwise destroy a gripper on the first run.
2. Check the model against the real cell
A simulation is only as good as its dimensions. Take a tape measure to the real fixtures, conveyors and guards, and compare them with the model. Small differences in height or position are the usual reason a perfect simulated path hits a real object. Update the model and note who is responsible for keeping it current.
3. Do an empty-cell run at reduced speed
With no product, no operator inside the fence and the speed turned down, let the robot perform the full sequence. Keep a hand near the stop button. Watch for unexpected motions, hesitations, and any place where the path swings wider than the simulation predicted. Increase the speed in steps only after each pass looks clean.
4. Test the stops, not just the starts
Press every emergency stop, open every guard interlock and break every light curtain, in order, while the robot is moving. Time how far it travels before halting. Confirm that the restart procedure is understood and that it does not skip any check. Nothing about this test is exciting, and it is the one you will be most grateful for.
5. Feed it dummy parts
Use scrap or placeholder parts to test the real pick and place routine. Try them in awkward orientations, slightly out of position and, if relevant, a little damaged. This reveals how forgiving your gripper, vision system or sensors really are. The way sensors feed a robot’s decisions is explained in this overview of how industrial robots combine sensor signals to act, which is a useful primer before you deliberately confuse them.
6. Rehearse the failure paths
Unplug a sensor, block a conveyor, drop a part, lose network connectivity for a few seconds. Record what the robot does and what message the operator sees. A good result is a safe stop and a clear instruction. A bad one is a silent stall that only an expert can diagnose. Fix the messages now, while the cell is quiet.
7. Walk through a shift change and a maintenance visit
Hand the cell to a different operator and have them start it using only the written instructions. Then simulate a maintenance task: lock out, change a tool, restore, restart. Pay attention to every point where someone had to guess. For mobile robots and warehouse fleets, the same exercise applies to charging, traffic conflicts and relocating a stuck unit, themes discussed in the longer walkthrough of driverless vehicles moving stock around a warehouse.
8. Time a full production-like shift
Finish with a longer run at normal speed using realistic loads and the real operators. Log stops, their causes and their durations. The figures will differ from the vendor’s brochure, and they are what you should plan production around. Write the findings into the cell’s handover pack so the next commissioning team can start from them.
Who should be in the room
A dry run goes better with the right mix of people. Bring the integrator or engineer who wrote the program, a safety representative who will sign off on the stop tests, an operator from each shift who will run the cell day to day, and a maintenance technician. Each notices different things. The programmer sees motion that looks wrong, the safety lead sees guards that can be bypassed, the operator sees awkward reach, and the technician sees parts that will be hard to service. Ask each of them to write down one concern before the first run and one afterwards, and compare the lists.
Resist the urge to invite spectators from management until the cell has passed the stop tests. Visitors create pressure to speed up, and speed is the opposite of what the early steps need.
Why the order matters
The steps move from virtual to physical, and from the robot alone to robot plus people. That sequence means each mistake is discovered at the lowest possible cost. A bad path is a line in a simulator, not a crushed tool. A weak stop procedure is found with scrap parts instead of with a person nearby. Skipping ahead trades that safety for speed, and the trade rarely pays.
One more habit is worth adopting: photograph or record each run. A short video of the first reduced-speed pass, kept with the commissioning notes, lets the team compare later behaviour with the original and settle arguments about whether a motion has changed.
A parallel from a very different workshop
Craftspeople of every kind trial their methods before the important job. That is the idea behind Blendz, an international archive of heritage cooking where readers can contribute recipes and learn the stories behind ingredients, where cooking methods are valuable because someone has already tested them. A commissioning checklist plays the same part on the factory floor.
Keeping the results useful
Store the dry-run log with the cell’s drawings and safety documents, and update it after any change to the program, the tooling or the layout. Repeat the stop tests on a regular schedule, because guards and switches wear. If you introduce a new part, run steps five and eight again at least in a shortened form. A dry run is not a one-time ceremony; it is how a cell stays understood over its working life.
Finally, share what you learn. A short note to the rest of the plant describing what surprised the commissioning team will save the next project a surprise of its own.
