Also searched as aluminium profiles for automation frames, automation equipment frame design, aluminium profile machine frame assembly, t-slot frame squaring, aluminium profile torque, modular frame assembly guide.

The error that grows with length

Ask anyone who has built a large modular frame what the hardest part was and they will not say the design. They will say getting it square. The reason is a small geometric effect that is invisible on a short member and enormous on a long one: an end face that is not perpendicular to the profile axis turns the member into a very long lever for its own error.

The table also explains why a bigger section feels easier to build square, and it is not only because of the width. A larger section usually arrives with a better end, because the cut is made on a bigger machine and the operator can see the error.

Two ways to remove the problem entirely. The first is to buy the members cut to length with a machined end rather than a mill cut: a mill cut is typically good to about ±1 mm with a burr, a machined end to about ±0.2 mm with a clean square face. The second is to use a connector that references the end face deliberately — an anchor connector is specified with an end-face squareness within 0.2 mm and a hole position within 0.15 mm, so the joint’s own accuracy sets the frame’s accuracy rather than the saw’s.

Square it before you torque it

This is the single biggest procedural difference between a frame that comes out true and one that does not. Torquing a skewed frame does not pull it straight; it locks the skew in and makes it permanent, because every joint is then clamped in a stressed position and the frame will move when the load comes on.

1. Assemble finger-tight only. Get every member in place with the bolts just nipping. The frame must still be able to move.
2. Put it on a flat surface and check the plane. A frame assembled on an uneven floor will be twisted, and no amount of diagonal measurement will find it. Use a surface plate, a large machined table, or shim the corners until the frame does not rock.
3. Measure both diagonals. On a 2,000 mm square frame, the two diagonals should agree within about 1 mm. That corresponds to roughly 0.35 mrad of skew, about 0.7 mm of error across the frame.
4. Correct by tapping, not by tightening. Bring the long diagonal in by a light tap at the corner with a soft mallet, then re-measure. Tightening a bolt to pull a frame square works once and then distorts the joint.
5. Torque in a pattern, starting at the most constrained corner. Work from the corner where the most members meet outwards, in stages — about half torque first, then full. This lets the frame settle evenly instead of building in a locked-in stress.
6. Re-measure the diagonals after torquing, and record them. If the frame has moved, you have found a joint that was not seated. Loosen, reseat, and start again from step three.

Torque, and the twenty-four hour re-check

Torque is a structural number, an electrical number and an assembly-sequence number at the same time, and the values are not interchangeable across series. Using a 40 series torque on a 30 series joint will deform the slot lips; under-torquing a 40 series joint leaves a connection that looks correct and is neither rigid nor conductive.

Then there is the step almost nobody takes. A bolted aluminium joint loses 5 to 10 per cent of its preload in the first 24 hours to embedment — the microscopic crushing of surface asperities under the nut head and in the thread. The joint has not failed; it has simply settled. But a joint that has settled is also a joint that has loosened, and on a frame that has to be both rigid and electrically continuous, a loose joint is two faults at once.

Leave access for the tool

A frame that cannot be assembled is not a design, it is a drawing. Every modular joint needs physical room for the tool that will tighten it, and the clearances are set by the slot, not by preference.

Feet, castors and the over-constraint problem

Support layout is where a good frame is quietly ruined, and it is decided by how many feet you fit rather than by how strong they are.

Four feet is statically determinate. The frame sits on three points and rocks onto the fourth when it is loaded. That is correct behaviour: the frame is not being bent by its own supports. Level the four feet, load the frame, and re-check.

Six feet is over-constrained. If the floor is not perfectly flat — and no factory floor is — the extra pair of feet either carries nothing or lifts the frame into a curve. The frame is then pre-loaded in bending before any process load arrives, and the diagonals you carefully set will drift.

Panels, guards and the vibration nobody predicted

Guarding and panels are usually treated as an afterthought, and they are the most common source of noise, rattle and warranty calls on an automation frame. Two rules cover most of it.

Retain the panel along its length, not at its corners. A 4 mm polycarbonate panel bolted only at four corners on a 1,000 mm span will drum at a low frequency, and it will do so intermittently depending on the process. Panel mounting strips in the 8 mm and 10 mm slot sizes spread the retention along the panel edge, which removes the drumming and stops the panel from being a point load on the frame.

Size the guard frame for the panel, not for the machine. The guard carries the panel’s weight and wind load and the door’s impact, not the machine’s process load. That is usually a much smaller load, and it is why a 30 or 40 series frame is normal for guarding where the machine base is 60 series or larger. The design drivers are the safety distances and the impact resistance, which are set by ISO 14120 and ISO 13857, and the door alignment, which is set by the hinge and the frame squareness you established earlier.

The corollary is worth stating plainly. A guard frame that is 10 mm out of square will show up as a door that catches, a latch that does not engage, and a panel that does not sit flat. The fault will be attributed to the hinges. It is almost always the diagonals.

Robot pedestals: it is rotation, not bolt strength

When a robot is mounted on a modular frame, the conversation usually goes straight to the bolts. That is the wrong first question, though it is not an irrelevant one.

The bolt check. An M8 T-nut in a 40 series slot carries 1.5 to 2.2 kN of pull-out. On a bolt pattern of 400 mm square — eight fasteners at 200 mm from the centre — four of them resist a moment about one axis, giving about 4 × 2 × 0.2 = 1.6 kN·m of resistance. A 6-axis robot with a 1,000 mm reach and a 10 kg payload produces a static moment of roughly 0.1 kN·m from the payload alone, and typically 0.5 to 1 kN·m once the arm mass and a dynamic factor for an emergency stop are included. So the anchorage is adequate, but it is not generous, and it should be checked rather than assumed.

The check that actually decides it. A rotation at the robot base becomes a positioning error at the tool, multiplied by the reach. A base rotation of 0.1 mrad — which is a very small angle, about 0.006 degrees — displaces the tool by 0.1 mm at a 1,000 mm reach. If the robot is specified to ±0.05 mm repeatability, the base has already thrown away the specification before the robot moves.

The site acceptance check, in six lines

What to record before the frame is signed off, and what to re-check at the first service. It takes twenty minutes and it converts a set of opinions into a baseline.

Then write the six results on the frame with the date, and re-run the same six checks at the first service. The value is not the individual numbers — it is the comparison. A frame that has moved 2 mm on its diagonals has told you which joint is failing before it fails completely.

Frequently asked questions