Quanti assi servono davvero? Il quarto si paga tre volte

A question emerges in every negotiation, as precise as taxes: How many axes do we need? Immediately after, the proposal that costs more than all others appears: Add one more, just in case. Those working in industrial automation have heard the phrase as an alarm bell for years: Adding an axis to a Cartesian robot is not an option on the shelf; it is a cinematic decision.

And cinematic decisions come with three different currencies: money, rigidity, and hours. What truly adds an axis? A rotating axis, when present, is almost always attached to the vertical axis, below the carriage Z, which rotates the gripping organ around the vertical. Its purpose is to change the orientation of the part between the gripper and the storage.

It serves only one purpose: to change the orientation of the part between the gripper and the storage. If the cycle does not contain rotation, the fourth axis adds no capacity; it adds weight. And weight on a carriage that accelerates and brakes hundreds of times at a time is not a detail. The rotating unit with its servo motor sits at the top of axis Z, in the worst possible place: it increases the inertia of the moving group, extends the setup time after each positioning, and asks the vertical axis dimension to bear weight that does not work.

On fast cycles, a few tenths of a second more setup time at pick time eats exactly the margin for choosing a Cartesian over an anthropomorphic. Rigidity does not forgive the joints granted. The reason a Cartesian robot can work well for twenty years is tedious: the beam is supported by two actuators, the load is distributed over the structure, the flexions are small and predictable.

Every added joint is a point where that predictability is crimped. A rotating axis introduces an additional coupling between the carriage and the gripper, with its play, torsional looseness, and the cabling that must follow the rotation - a dimensioned cable chain, connectors that work, an additional actuator to tune and keep in service.

Nothing dramatic, if the axis works. But if it does not work, you have paid for a more flexible, slower setting machine, and more expensive to maintain for the assurance against a cycle that does not exist. The assurances in mechanics pay for every intervention over twenty years. There is also the test, which has its own times and does not appear suddenly.

A rotating axis must be put in place along with the other three, its final test must be verified for safety, and its emergency behavior must be tested: what does the gripper do if the fourth axis stops halfway through a rotation with a part in hand? These are questions that in testing are paid for in days, not minutes - in our experience, on a medium-sized Cartesian, an extra axis easily adds half a day to testing and a day to programming and cycle tests before the first good piece.

And there is the homing: after every serious shutdown, that axis must find its zero repeatedly, otherwise the first layer of the pallet will be rotated by two degrees and you will notice it on the third. Sometimes only two axes are enough, and it is not a retreat. The discussion over the fourth axis hides the more useful one about the second and third.

A surprising number of line-end cycles live on only two axes: picking at a fixed height from a conveyor, storage at a fixed height on a pallet that descends on an elevator. There a two-axis dimensioning - X for the long run, Z for the height - is the right machine, and the third axis would be a gift. Conversely, where the heights of picking change with each cycle or the pallet is built layer by layer without an elevator, it is the Y axis or generous longitudinal run that makes the difference between a fluid cycle and one that waits.

The healthy hierarchy is this: first, set the run, carriage, and speed numbers based on the true numbers of the cycle; then count the axes those numbers require; the rotation comes last, and only if the cycle calls for it. The alternative that no one draws is whether rotation must be forced on the gripper? Often the answer is no. If the part needs to be rotated between stations, a ground-level rotating table, a reorienting roller on the conveyor, or the load scheme itself can rotate.

Moving the rotation from the tip of axis Z to a fixed ground actuator means removing weight from the mobile group, simplifying the cabling, and leaving the Cartesian with three axes to do what it does best: move straight, fast, and rigid. It is not always possible - when every jaw needs to be oriented differently depending on the layer, rotating at the gripper is the right choice - but it must be compared, not taken for granted.

The test I use to close the discussion is brutal and does not require software: describe the cycle without that degree of freedom. If the cycle stops, either to make it turn, or to make it happen requires an extra person on the line, the axis is justified and dimensioned with a margin. If the cycle runs the same, with the same times and quality, that axis you are buying to sleep peacefully.