A robot cell is a stiffness and damping problem before it is a strength problem

The loads a robot puts into its frame are modest. A mid-size industrial arm might apply a few hundred newtons at the tool, and the reactions at the base are well within the capacity of a 45 or 60 series profile. What a robot does apply is motion, and motion is where a lightweight frame behaves differently from a welded steel bench.

Two parameters matter. The first is static stiffness: how far the tool deflects under a given force. Static stiffness sets the accuracy of the cell under a steady load, and it is dominated by the base plate and the first metre of structure, because deflection accumulates from the floor upwards. The second is dynamic behaviour: the frame has natural frequencies, and if the robot excites one of them the tool vibrates. Vibration shows up as a positioning error, as a longer settling time, and therefore as a longer cycle.

Aluminium is stiff, but it is not as stiff as steel for the same section: the elastic modulus is around 70 GPa against about 210 GPa. What aluminium has is a much better stiffness per kilogram, so for a given mass budget a well-designed aluminium section can be deeper and therefore stiffer than a steel one. The design consequence is that a robot cell frame should be sized by a deflection limit at the tool, not by a stress check, and that depth and the base plate do most of the work.

Aluminium has no fatigue limit, and a robot loads its frame millions of times

Steel has a fatigue limit: below a certain stress range, a steel detail can be cycled indefinitely and the S-N curve flattens out. Aluminium does not. Its S-N curve keeps falling, so there is no stress range below which a cycle is free. Every cycle of a robot move takes a little life out of the detail that carries it.

In practice this does not mean aluminium is unsuitable; it means the design has to be done on stress range rather than on peak stress, and it means the details matter more than the section. A fatigue crack starts where stress concentrates: at a sharp change of section, at the toe of a weld, at the root of a thread, and at a slot lip that is being point-loaded. A bolted joint that is properly preloaded does not fret and does not fatigue at the bolt hole, because the clamp force carries the load as friction. A joint that is loose rocks, frets, and initiates a crack.

The controls are ordinary. Use brackets that spread load into the profile instead of loading a slot lip. Keep the number of joints in the load path small. Preload fasteners to the manufacturer’s torque for the actual friction condition. And where a robot applies a genuinely cyclic load to a cantilever, check the stress range rather than the static stress, because the static check will pass.

The safety standard changed in 2025: ISO 10218-1 and ISO 10218-2

ISO 10218-1 and ISO 10218-2 were published in new editions in 2025, the first major revision of the industrial robot safety standards in well over a decade. Part 1 addresses the robot itself and Part 2 addresses the robot system and its integration, which is the document a cell builder works to.

The significance for a frame supplier is that the boundary between the robot and its installation has become more explicit. Functional safety requirements, the treatment of collaborative applications, and the expectation that the integrator performs and documents a risk assessment all sit in Part 2. A frame that is described as a structure and a fence that is described as guarding are no longer separate conversations: the assembly has to be assessed as a system, and the safety-related parts of the control system have to be designed to a performance level.

For a buyer, this changes what should be in the enquiry. Instead of asking for a frame with guarding, the enquiry should say which safety functions the structure carries, what the required performance level is, and who is producing the risk assessment. A profile supplier can provide the structure and the guard frame; it cannot certify a safety function.

The EU machinery rules change in 2027 as well

In the European Union, Regulation (EU) 2023/1230 replaces the Machinery Directive 2006/42/EC and applies from 20 January 2027. It is a regulation rather than a directive, so it applies directly without national transposition, and it modernises the framework for digital instructions, software, and machinery that includes artificial intelligence or autonomous functions.

For a robot cell built on an aluminium frame, the practical effects are documentation and traceability. Technical documentation, the declaration of conformity, and the instructions have to be maintained for the life of the machine and produced in a digital form where required. If the frame carries a safety function, that function is part of the machinery and is inside the assessment. A cell that is assembled on site from a kit is still a machine, and the person who places it on the market is the manufacturer for the purposes of the regulation.

None of this changes the profile. It changes the paperwork that has to accompany the assembly, and it is worth knowing before a cell is designed, because a safety function that is discovered late is expensive to add.

A T-slot frame is a good structure and a poor safety fence

This is the most common confusion in a robot cell specification. A T-slot frame is an excellent structure and an excellent carrier for panels. It is not, by itself, a guard.

A guard has to contain a hazard, and that means it has to have a defined impact resistance, a defined opening size that keeps a person out, and a defined fixing that will not pull out. A slotted aluminium post with a thin panel in it may satisfy none of those, and a slot that accepts an M8 bolt is also an opening. The guard design has to follow the machine guard standard for the hazard, with the panel thickness and the fixing chosen for the impact, and the frame then sized to carry that panel.

There is also a serviceability point. A guard that is bolted into a slot can be removed in a minute by anyone with an Allen key, which is convenient during commissioning and dangerous during production. Where a guard is a safety-related part, its fixing should be tamper-resistant or interlocked, and the drawing should say so. The same frame can be both, but only if the intention is written down.

Anodising and the electrical problem

An anodised aluminium profile has an oxide layer that is an electrical insulator. That is an advantage for corrosion and a problem for anything that needs continuity: equipotential bonding, ESD control, and any sensor or cable tray that relies on the frame as a return path.

So a robot cell frame that is bonded has to be bonded deliberately. Bonding washers that bite through the anodic layer, dedicated conductors from part to part, and a measured continuity check at each joint, are the normal controls. The mistake is to assume that because the frame is metal it is continuous; the anodic layer is on every face that was anodised, and the joints are the places where continuity is least reliable.

The same point applies to a painted or powder-coated frame, and to a frame that mixes anodised and mill-finish parts. The bonding plan has to be a drawing, not an assumption, and it has to be verified after assembly, because continuity that is lost when a joint is re-tightened is continuity that will be lost in service.

Thermal drift and the long rail

A robot cell is often built in a workshop that is warm in summer and cold in winter, and a robot cell is a metrology problem. Aluminium expands by roughly 23 micrometres per metre per kelvin, so a six-metre rail that sees a 15 kelvin swing moves by about two millimetres. On a cell that is positioning to a tenth of a millimetre, that is a real error, and it is invisible in a static drawing.

Three controls are available. The first is to anchor the rail at one point and allow the rest to slide, so the rail grows away from the datum rather than fighting its fixings. The second is to keep the rail in a controlled-temperature enclosure, or to compensate in the control system. The third is to accept the drift and design the tolerance budget to include it. What does not work is to bolt both ends of a long rail rigidly to a stiff floor and expect the rail to stay straight: the movement goes somewhere, and it usually goes into the fixings or into a bow.

The same calculation applies to the base plate and to the mounting of the robot itself. A robot mounted on a plate that is restrained at both ends is a robot mounted on a plate that is being bowed by its own expansion.

What to send with a robot cell enquiry

The information that makes a robot cell quotation useful is not the profile list. It is the robot, the payload, the reach and the accuracy the process needs; the deflection limit at the tool; whether the cell is a fixed or a mobile platform; which safety functions the structure carries and to what standard; the electrical continuity requirement; and the temperature range the cell will see. Given those, the section and the base plate follow.

It also helps to say who is the integrator. The frame supplier builds the structure, the integrator owns the risk assessment and the safety functions, and the boundary between the two is the single most useful thing to write down before the order.