Single points of failure: the machine that can stop your whole site

It's a Wednesday in November. The site is busy. Twelve lines running, two shifts of pickers in, trailers queued on the yard.

Then the compressor trips. The standby was taken out last month for a bearing and never put back. Within ten minutes every pneumatic gate, cylinder and gripper on site is dropping out. Twelve lines, stopped by one machine that nobody thinks about because it lives in a shed round the back.

Every site has one of these. The shrink tunnel every line feeds through. The single sorter induct. The one boiler. The wrapper at the end of the only dispatch lane. The bit of kit that, when it stops, everything stops.

The trouble is that most sites find out which machine it is on the day it fails.

Looking in the wrong place

Ask a production manager which machine matters most and they'll name the big one. The fastest line, the newest filler, the one with the biggest order book.

Those matter. But a fast line that stops is one line stopped. A single point of failure is the thing many lines depend on. It's often small, old and boring. It sits in a plant room or at the end of a conveyor, and it gets the least PPM time because it "never goes wrong".

Here's the test we'd use. For each machine, ask one question: if this stopped now, what else would stop with it? Write the answer down. The machines with the longest answers are your single points of failure.

Where to look

Walk the site with that question in mind. These are the places it usually hides:

• Utilities: compressed air, steam, chilled water, the main incomer and the panel that feeds the whole line. Our post on compressed air down goes through one of the worst.
• Merge points, where several lines feed one machine. A shared palletiser, a single check weigher, one sorter induct, one spiral to the mezzanine.
• End of line: the stretch wrapper or the label printer on the only dispatch lane. Nothing ships without it.
• Control: one PLC running three cells, one server running the WMS, one network switch in a cupboard nobody can find.
• People: the one engineer who knows how the old sorter works. More on that below.

Don't skip the boring kit. The chiller that keeps a cold store at temperature is a single point of failure even if no line touches it.

Rank it by what it costs

Once you have the list, put a number on each one. Work out what an hour stopped costs: lost output, idle labour, overtime to catch up, product at risk, and any penalties for late delivery. Our post on what an hour of downtime costs walks through the sums. For a single point of failure, multiply by every line it takes down with it.

Then think about how long a breakdown would really last. A sensor fault on a wrapper might be twenty minutes. A burnt-out motor on a compressor with no spare could be three days. The machines that combine a big hourly cost with a long likely repair go to the top. That ranking is the start of a proper criticality ranking for the whole site.

Cover it before it fails

You have four ways to deal with a single point of failure. Most sites need a mix.

Remove it. A second compressor on auto-changeover. A bypass lane round the wrapper. A manual way to run product past a dead machine. This costs money, but for the worst one or two it often pays back the first time it's used.

Maintain it harder. HSE's code of practice for PUWER says that where safety-critical parts could fail, "a formal system of planned preventative or condition-based maintenance is needed" (L22, paragraph 75). That's about safety, but the same thinking works for uptime. Put your single points of failure on PPM first, and add vibration, thermal or oil checks where the failure would hurt most. HSE puts the link simply: an effective maintenance programme makes plant more reliable, and "fewer breakdowns will mean less dangerous contact with machinery" (HSE, maintenance).

Hold the spares. For each one, list the parts that would turn a three-day stop into a three-hour one. Motor, gearbox, drive, PLC card, the odd bearing nobody stocks. Then check they're on the shelf, not on a spreadsheet. Our post on critical spares covers what's worth holding.

Line up the people. This is the one sites miss.

The engineer is a single point too

A machine with a spare motor on the shelf can still sit stopped for a day if nobody on site can fit and commission it. The person who knows the drive settings is off. The controls engineer who wrote the code left two years ago. The night fitter is mechanical and the fault is in the panel.

So for each single point of failure, ask who can fix it. On days, on nights and at weekends. If the answer is one name, you have two single points of failure, not one. Our post on single-skill risk goes into building cover around that person, and when the only person who knows the machine is on holiday covers the fortnight they're away.

Then ask the harder question. When that machine goes at 2am on a Sunday, who do you ring? If the answer is "we'd work it out on the night", you'll lose the first two hours to finding someone. Work it out now. Know which outside engineer, with which skills, can get to your site and how fast. Have the machine details, drawings and backups ready to send them.

Here's our view. For your worst single point of failure, the people plan matters as much as the spares. A spare motor with nobody to fit it is just stock.

Write it on one page

For each single point of failure, put this on one sheet and keep it by the machine and in the engineering office:

1. What it is, and what else stops when it stops
2. Cost per hour when it's down
3. The spares held, and where they are
4. Who can fix it, by shift, and their numbers
5. Who to call from outside, and what to send them
6. Any manual workaround, and who can approve it

When it goes, nobody should be searching for any of that.

What to do on Monday

1. Walk the site and list every machine where the answer to "what else stops?" is more than one line.
2. Put a cost per hour on the top three, and check their spares are on the shelf.
3. For each of the three, write down who can fix it on every shift. Mark any with only one name.
4. Agree who you'll call from outside for each one, before it fails.

* Paragraph numbers refer to L22, fourth edition (2014, amended 2018).