I do a lot of these site visits. Maybe 40 last year, give or take. And honestly, the same five problems show up in different combinations almost every time. None of them are exotic. None of them need a structural engineer. You just have to know what to look for.
So here’s the list, in roughly the order of how often we see them.
Jump to
1. The wall is thinner than you think
Everyone specs “28 mm pipe.” Almost nobody specs the wall thickness. That’s the whole problem.
The outer diameter doesn’t do much for stiffness. The wall does. Drop from 1.5 mm wall to 1.2 mm and you’ve lost about a fifth of your bending stiffness for what looks like the same tube. The rack doesn’t fail. It just starts to feel “a bit bouncy.” Six months later you have a 22 mm sag.
The formula isn’t a secret. Deflection at the centre of a simply-supported beam is:
What that I term hides is everything. For a hollow tube it’s π(D⁴ − d⁴)/64, and that inner diameter is where the wall thickness shows up. Thinner wall means a bigger d, means a smaller I, means more sag. Always.
What we actually use on the shop floor, in plain language:
| Tube | What it’s good for | Where it fails |
|---|---|---|
| Ø28 × 1.0 | Label holders, signal lights, sensor brackets. | Anything heavier than a clipboard. |
| Ø28 × 1.2 | Empty totes, kanban cards. The cheap stuff. | Loaded shelves. Sags within a year. |
| Ø28 × 1.5 | Picking flow racks, loaded totes up to ~5 kg each. | Heavy bins, anything past 30 kg per metre of span. |
| Ø28 × 2.0 | Tool trolleys, fixtures, anything that gets leaned on. | Rarely fails. Often overkill. |
| Ø42 × 1.7 | Main verticals on tall racks. Pallet flow. | Costs more and weighs more. Worth it for verticals over 1.8 m. |
If you’re writing a spec for a new rack, write “28 × 1.5 minimum” on the drawing. That one line will save you a callback.
2. Snap-on joints where you needed compression joints
In a catalogue photo, the two look identical. On site, you can spot them in about three seconds.
A compression joint has a hex bolt. You tighten it to maybe 8 or 10 newton-metres and the joint locks onto the tube. Properly torqued, one of these will hold 100 kg before it slips.
A snap-on joint has a spring clip or a plastic detent. No bolt. You push the tube in until it clicks. The whole point is reconfigurability — you can pop the kanban arm off the rack in two seconds. The price you pay is that it’ll hold maybe 8 to 15 kg before the tube starts walking out.
That walking-out is the silent killer. Snap-on joints don’t fail. They creep. Every time someone leans on the shelf, the tube slides a tenth of a millimetre. Do that a few thousand times and the shelf is visibly tilted. Operators just learn to push it back into position every morning until one day it doesn’t go back.
So the rule is short: if two load-bearing tubes meet at a joint, that joint needs a bolt. End of story.
| Joint code (Yusi catalogue) | Type | Roughly holds | Use for |
|---|---|---|---|
| HJ-1 | 3-way compression corner | ~110 kg | Structural corners. The workhorse. |
| HJ-7 | 4-way compression cross | ~140 kg | Mid-span supports where shelves cross verticals. |
| HJ-12 | Parallel reinforcer (compression) | ~180 kg | Doubling a horizontal beam when you don’t want to upsize the tube. |
| HJ-22 | 3-way snap-on | ~8 kg | Label arms, signal-light brackets. Not structural. |
| HJ-30 | Plastic clip-on | ~3 kg | Cable runs, hose guides. Cardboard sleeves. |
Swapping a snap-on for a compression joint takes maybe a minute and a half. It’s the highest-ROI fix on this list and the one customers resist the most, because the snap-ons are already installed and they look fine. They’re not fine. Just swap them.
3. The span is too long (this one is brutal)
Span shows up in the deflection equation as L to the fourth. Not L. Not L squared. L4.
Read that again. Because it means doubling the span doesn’t double the sag. It multiplies it by sixteen.
900 mm span → 5 times the sag
1200 mm span → 16 times
1500 mm span → 39 times
This is why a rack that worked perfectly at 800 mm spacing becomes a hammock the day someone moves the verticals out to 1100 mm to make room for a new pallet jack. Nothing else changed. Same tube, same joints, same load. And it sags eight times more than it used to.
What we use as a working rule:
- Live load on Ø28×1.5: 900 mm between uprights, hard limit.
- Just structural weight, no product: 1200 mm, fine.
- Anything past 1200 mm: either go to Ø42 or stick an extra vertical in. Don’t fudge it.
- Cantilever arms (only one end supported): 350 mm max. Cantilevers are unforgiving.
If the rack is already built and over-spanned, you have three options. Cheapest is dropping in an intermediate vertical — one extra post, two HJ-7 joints, and you’ve cut the sag by 84% in twenty minutes. Next is doubling the horizontal with a parallel tube clamped on with HJ-12s every 250 mm. That roughly doubles stiffness. Most expensive is re-doing the whole horizontal in Ø42, which is only worth it if you also need to add a tier above.
4. Nothing is bracing it sideways
A pipe-and-joint frame, without a diagonal, is just a parallelogram. Push on it sideways and it leans. The only thing stopping it is friction in the joints, and friction is the first thing to go.
That’s the “leaning rack” you’ve probably seen on a shop floor — the one operators push back into position every morning. It’s not tired. It’s under-braced.
| Width / depth of the frame | What it needs |
|---|---|
| Under 800 mm | Usually fine without one. Joint friction holds it. |
| 800 to 1500 mm | One diagonal per open face. Corner to corner. |
| 1500 to 2400 mm | Two diagonals per face. K or X pattern. |
| Over 2400 mm | Split it into two bays with a middle vertical, then brace each bay. |
Two things people get wrong here. First, they land the diagonal on a snap-on joint — useless, because the diagonal is in tension and the joint just slips. Diagonals always land on a compression joint.
Second, they brace the sides and leave the back open “for access.” The back is exactly the direction the rack wants to fall — toward the operator. Brace it. Or anchor it to the wall. Don’t leave it open and hope.
5. The load is moving, and the design assumed it wasn’t
This is the expensive one. Usually because nobody notices it for three months.
The designer added up the static weight of all the totes, picked a tube, picked a joint, signed it off. Fine. What they didn’t add up was the dynamic stuff — an operator dropping a 4 kg tote from 200 mm up, an AGV nudging the front upright every time it docks, a roller track running 600 cycles a shift. Those events don’t add weight. They multiply it.
Roughly what we multiply by:
| What’s happening | Multiply the static load by |
|---|---|
| Operator places a tote down gently | 1.2 |
| Operator drops a tote from ~200 mm | 2.5 |
| Tote slides down a 3° flow lane and hits the end stop | 3 to 4 |
| AGV makes contact while docking | 5 or more |
| Forklift accidentally bumps the rack | 10 or more. Add a steel bollard. Pipe and joint is not a crash barrier. |
One customer story, because it’s exactly the failure mode you want to avoid:
Picking rack. AGV-fed. Designed and built correctly for static load. Ran for 14 months. Started dropping totes. We came out, tore it down, found the compression joints had walked 4 mm down the tube on the AGV-facing side. The AGV had been making contact 28 times per shift, three shifts, six days a week. About twelve thousand impacts. The rack hadn’t failed. It had just slowly walked itself apart. A £30 bollard at the dock would have stopped it.
So if your rack has anything moving near it — operators, AGVs, sliding totes, forklifts — check four things:
- The end stop on every flow lane. If it’s a single Ø28 horizontal, that’s wrong. Use a steel angle or double the tube.
- The first 250 mm of every flow lane. Where each tote impacts the one in front of it. Reinforce the rails here.
- The floor anchors at any AGV dock. M10 chemical anchors, four per upright, into concrete that’s actually thicker than 100 mm. Expansion anchors in thin slab are a waste of time.
- Re-torque every compression bolt on a dynamic rack every three months. Mark each bolt with a paint stripe so loose ones jump out at you.
A short checklist you can run today
Hand this to whoever does your rack inspections. It takes about 20 minutes per rack.
- Caliper-check three random tube ends. Anything under 1.2 mm wall, flag it.
- Walk the rack and find every load-bearing joint. Any snap-ons there? Swap them.
- Measure the span between uprights on horizontals that carry product. Anything past 900 mm, flag it.
- Push the rack sideways from behind. More than 5 mm of movement means it needs another diagonal.
- Spot-check ten compression bolts. They should be 8 to 10 N·m, with no thread showing above the nut.
- Anything taller than 1.6 m or carrying over 100 kg has to be floor-anchored. No exceptions.
- Lay a string line across each shelf. More than the span divided by 200 (so about 4.5 mm on a 900 mm shelf) means it’s already past serviceable. Redesign, don’t re-tighten.
That’s really it
Sagging racks aren’t a materials problem. They’re a spec problem, almost always one of the five above, often two or three stacked together. Each individual fix is cheap — a £3 joint, a couple of metres of tube, a floor anchor. The cost only shows up when a shelf goes onto a finished-goods tote at three in the morning.
If you’re looking at a rack right now and not sure which of the five is biting you, send photos and rough dimensions to [email protected]. We’ll come back with a load calculation and a parts list, usually same day.