Robotics and industrial equipment

How heavy a robot is, how accurate it is and where it can work are all set by its structure.

Structurearms, links, beams, frames, housings

Processeswinding / pultrusion / HP-RTM / moulding

Sized onallowable tip deflection, first frequency

Interfacesoriginal holes and mounting faces

Robot structures have always been metal. Move them to composites and weight, stiffness and sealing all shift at once; if any one of the three gives way, machine performance goes with it. That is the real barrier in this sector, and it is what this page sets out.

We supply arms and automation equipment, mobile chassis, special-purpose machines, and humanoid and quadruped platforms. Arms and links are filament wound, beams and frames pultruded, housings and covers made by HP-RTM or compression moulding. Forming and finishing sit in the same plant. Interfaces follow your drawing, and your assembly process does not change.

Where the weight comes off, the saving shows up

Machine weight divides into the part that stays still and the part that moves. They pay back differently.

The moving part pays back most. Servos are sized at full reach with the payload on. At that moment the weight of arm and frame sits entirely on the lever arm. That is before the payload is counted, and most of the torque is already spent. Take weight off the moving parts and the gearbox can drop a size, the base needs less ballast, the time to reach speed shortens.

Battery machines, tracked, wheeled, AGVs, and the arms mounted on their chassis, carry themselves with every move. Structure accounts for a large share of machine weight, so the saving turns straight into range: longer running on the same pack. The same range on a smaller pack with a shorter charge. Machines on swappable packs need fewer of them. Humanoid and quadruped platforms feel this hardest — drop the structure a step and the joint motors and their cooling both relax.

None of it holds if stiffness drops. Cut weight and lose stiffness, and tip vibration cancels what the smaller drive bought you; it shows up in cycle time and in scrap. So the question before any weight target is whether stiffness survives.

Equal stiffness, not equal strength

The commonest mistake when swapping metal for composite is designing to equal strength. The strength check passes, and the part turns out too soft once assembled. A machine tool does not stop chattering because the strength is adequate, and a robot does not stop vibrating because nothing has broken.

The order is the other way round. Set the allowable tip deflection, or the first natural frequency, and back-calculate the layup and wall thickness from it. The weight saving falls out of that calculation.

Keeping the interface unchanged matters as much. Hole positions, mounting faces, bolt sizes and assembly sequence stay as designed. Your tooling and process do not change — only the body does. On production equipment this matters more than a few percent of weight: move one hole and everything downstream needs re-validating.

Arm tubes and links are filament wound. Wall thickness need not be constant along the length: bending moment peaks at the root, so it can thin towards the tip. Taper is not cosmetic — it moves mass from the tip back to the root. On a moving part that transfer is worth more than weight lost overall. The sharper the taper the more the fibre slips, so the layup angle has to be recalculated for the taper; a straight-tube sequence cannot be reused.

Beams, frames and chassis are long parts, pultruded in I-section or square tube, with modulus set through layup and fibre volume fraction. Rail mounting faces and flange faces are machined on five axes, flatness to drawing.

Beam deflection does not scale linearly with span. Under a point load it follows the cube of the span; under a distributed load, close to the fourth power. Double the span and the same section deflects 8 to 16 times as much. Past a certain length the limit is therefore not strength, there is usually strength to spare. Stiffness: the tip is waiting for the structure to settle. What composites win here is specific stiffness, not strength.

One more point is easily missed: shop temperature moves with the seasons, metal beams expand, and repeat positioning drifts. Composite axial expansion is low, and with a symmetric layup this term stays small.

Joints

The joint is the weakest point in parts like these. The tube itself rarely fails; failure collects in the transition at the joint, where the fibre is interrupted and stress concentrates.

So we use adhesive bonding and mechanical fastening together. The bond carries the continuous load, the fasteners hold the peel end; each has one weakness on its own — pure bonding peels, pure bolting delaminates at the hole.

Inserts are embedded during forming and holes are made in one pass. We do not drill afterwards and drop in an aluminium block: a hole cut after cure severs fibre. The hole edge becomes a crack starter.

Carbon conducts. Connected straight to metal, the contact face corrodes galvanically in damp or salt conditions, and the metal side goes first. So joints are isolated — glass-fibre washer or sealant, and doing both is no excess.

Housings and covers: sealing is settled at the moulding stage

Sealing is won or lost at forming, not at assembly. The gasket is the last item in the chain. How far it compresses and where it escapes is set by groove geometry and the hard stop — both fixed in the tool.

This one came up on a special-purpose wheeled robot. The machine was designed on a two-year service cycle: strip, replace the seal, apply sealant on site. After each service the leak rate was still 15%.

The problem was not the number of services but the operation itself: bond-line thickness, surface cleanliness, cure conditions and operator technique. On site none of the four closes the loop, and there is dust on top of that. The gasket itself came out intact.

We moved the sealing face into the plant, formed in one shot, with compression held at 20–30% by a hard stop; nothing is opened on site now beyond fastening and visual checks. At the two-year inspection the leak rate was down to 3%, and downtime per service from 4 hours to 50 minutes.

The rest of the housing is settled at the same stage:

Static displacement plot of a tracked-robot upper housing
Static displacement plot of the upper housing: red marks peak deflection under top-face load, which is what sets where the reinforcement goes.

Top-face loading uses local thickening or a sandwich, never overall thickening — overall thickening adds weight and cure distortion together.

Warpage: a temperature difference between mould halves and an asymmetric layup both bend the part on demould. No amount of later machining recovers it. Tool and layup have to be designed together.

In harsh service — dust, washdown, chemical media — the surface layer and coating are fixed at forming. Protection is built into the structure, not applied afterwards.

The gasket is rarely the culprit. Groove constraint has a page of its own: a seal that looks perfect and still leaks usually sits in a groove that holds only half its section — from single-sided to two-sided.

Robotic seals fail in two places in practice. One is the groove above. The other is the enclosure itself: it flexes under load, the joint opens, and the gasket is blamed for a stiffness problem — 20 kg lighter with IP67 kept, and leak rate 15% to 3% in one redesign.

Electrical requirements decide the material

Carbon conducts, glass insulates. The choice is made when the material is picked, not in post-processing.

Where stiffness is needed, carbon; where electrical isolation is needed, glass. The two can share one part: carbon carrying the load, glass providing the isolation, formed together in one shot so there is nothing to assemble.

Control-box housings do not carry the earth path; earthing goes through a separate metal part. The housing itself is not shielded — where shielding is required, a conductive coating or metal liner is added to your specification. Insulation class and other electrical requirements are selected to the standard you give us, not assumed.

What we need before quoting

What decides the quote is the duty rather than the drawing: moving mass, maximum reach, allowable tip deflection or first frequency, temperature range. Cleaning media, ingress protection. Whether the machine is fixed or battery-powered.

Annual volume and delivery date decide the process route and tooling spend; appearance and paint specification decide the finishing steps. Without them the number can only be an estimate.

Process by part type: arms and links wound, beams and frames pultruded, housings, end plates and flanges by HP-RTM or compression moulding, inserts embedded at forming.

Finishing stays in the plant: five-axis machining, drilling, tapping, painting, assembly and leak test. Nothing is subcontracted, so the line of responsibility is clear.

Parts we have made in this sector

Grain levelling robot

Phosphine fumigation: the copper heat path was ruled out by the medium, heat now travels through the structure. Peak internal temperature fell from 85 °C to 65 °C.

What we do not take on

  • We do not build whole machines, and we do not do electronics.
  • We do not take weight-only jobs that ignore stiffness. Such a part shakes on the machine, and the conversation comes back to design.
  • We do not machine customer-supplied parts. Finishing applies to parts we formed ourselves; that is where the line of responsibility sits.

FAQ

Which parts on a robot can move to composites?

Arms, links, beams, frames, chassis, housings, covers, control boxes, the limbs, frames and joint housings of humanoid and quadruped platforms. The brackets and connectors around them. The test is not whether it can be made, but whether stiffness, sealing and interfaces survive the swap.

Can a carbon part replace an aluminium one directly?

Not on equal strength. Aluminium is selected on yield, composites on allowable deflection or first frequency, and the weak points differ. Keeping the original mounting interface usually means recalculating the layup and thickening around the interface.

Why is not every robot built from composites?

The cost has to answer to the return. On a fixed machine with a small payload and a short stroke, aluminium is enough. The drive saving does not cover the price of the part. Long reach, high speed, tight tip accuracy, or a battery-driven mobile machine — there it works.

Do you take small batches?

Yes. Tooling follows volume: a temporary tool for prototypes, steel for series. There is no need to cut a tool before the volume is settled.

What accuracy can you hold?

Machining to the tolerances on your drawing, on five axes, to the same standard as our product pages. In composites the difficulty is cure distortion, not machining. That has to be held in the tool and the layup. Machining cannot put it back.

Is there warning before a composite part fails?

Not the yield plateau metal gives you. Metal yields before it breaks, and you can see the deformation. A composite has no yield stage and deforms little before failure. Appearance tells you almost nothing. Design is therefore governed by allowable strain rather than ultimate strength, and equipment working near people carries an added safety factor. That is also why a quote cannot be read off weight alone — how the safety factor is set changes the layup.

Can a housing be supplied with the seal in place?

Yes. Groove and gasket seat are formed in one shot, leak-tested before dispatch. What arrives is an assembled housing rather than a kit of parts.

Send the duty first, the drawing second. Once the duty is fixed, the process and the tool follow.

Send the duty

sales@atomfrp.com WhatsApp +853 68617664

Send the drawing. We will send the route.

A drawing, a photo of a failed part, or three lines of requirement is enough to start. Scope, price indication and lead time within 24 hours.

No commitment · NDA available · Reply within 24 hours

Request a quote WhatsApp