In short: Condition monitoring, putting vibration, temperature and acoustic sensors on machines so failures announce themselves weeks in advance, is the most proven Industry 4.0 investment a UK factory can make, with 30-50% reductions in unplanned downtime routinely reported. Yet most plants never get past a ten-sensor pilot, and the blocker is rarely the sensors or the analytics: it is the cost and fragility of getting data off thousands of retrofit points on a brownfield factory floor. That is a network problem, and it is exactly the problem private 5G was designed for.
Key Takeaways
- Machine downtime is the cost that hides in plain sight — an hour of unplanned stoppage on a production line runs from thousands to tens of thousands of pounds, and most UK plants still discover failures when the machine stops rather than weeks before.
- Pilots succeed, rollouts stall — ten sensors on one line prove the value quickly, but scaling to the thousands of points a whole plant needs multiplies wired costs per point and overwhelms factory WiFi, which is where most condition monitoring programmes quietly die.
- Wireless-first retrofit changes the arithmetic — battery-powered sensors over one licensed private 5G network cut the cost per monitored point dramatically and reach the presses, pumps and packaging lines that were never going to justify a cable run.
In a nutshell
What is condition monitoring, and what is it worth?
Condition monitoring is the practice of instrumenting machines so their health is measured continuously rather than inspected occasionally: vibration sensors on bearings and gearboxes, thermal sensors on motors and drives, acoustic and ultrasonic sensors for leaks and electrical faults, current signatures on the panel. The point is to catch the slow drift that precedes failure, a bearing running two degrees warm, a vibration signature that has changed shape, weeks before it becomes a seized spindle at 2am on a Saturday.
The value case is unusually well evidenced for an Industry 4.0 technology. Across published deployments the consistent findings are 30-50% reductions in unplanned downtime and maintenance cost savings in the 10-25% range, with the added, quieter benefit that planned interventions happen at planned times with parts on the shelf. For a UK plant, the arithmetic starts with an honest number for what an hour of unplanned stoppage costs: on a food line feeding a retailer's depot window it is measured in missed orders and service-level penalties; on an automotive or aerospace line it runs to tens of thousands of pounds per hour. Most maintenance managers know this number viscerally even where finance has never written it down.
We would add one candid qualification: condition monitoring is not magic. It catches the failure modes that develop gradually, which is most mechanical ones, and does little for the genuinely sudden. The plants that do well treat it as a better information supply for an existing, disciplined maintenance operation, not a replacement for one.
Why do condition monitoring projects stall after the pilot?
The pattern is so common it deserves naming. A plant picks its worst-behaved line, fits ten or twenty sensors, catches a failing gearbox within three months, and the pilot is declared a success. Then the rollout plan lands: the full site needs two, five, ten thousand monitored points across presses, pumps, compressors, fans, conveyors and packaging machines, many of them decades old, none of them built with instrumentation in mind. And the programme stalls, for reasons that have little to do with sensors or software.
Firstly, the wiring economics collapse. A wired sensor point on a live factory floor, cable tray, conduit, isolation, and the production interruption to install it, costs far more than the sensor itself; multiplied across thousands of points it turns a compelling business case into an unaffordable one. This is the same trap that greenfield plants are now avoiding by designing cable trays out entirely, but a brownfield plant does not get to start again.
Secondly, factory WiFi cannot carry the load. Wireless sensors solve the cabling problem and immediately meet the factory floor's radio reality: metal everywhere, interference from drives and welders, and unlicensed 2.4GHz spectrum already crowded with scanners, AGVs and a decade of accumulated devices. A few hundred sensors chirping politely is fine; thousands of endpoints, some streaming high-frequency vibration waveforms for detailed analysis, is a density and reliability problem WiFi was never engineered for. We have covered the WiFi versus private 5G decision in detail; condition monitoring at plant scale is one of the cleanest cases where the licensed option wins.
Thirdly, ownership goes missing. The pilot was maintenance's project; the rollout needs IT (network), OT (controls), production (access windows) and finance (a multi-year case) to agree. A surprising number of programmes die not from any technical failure but because nobody owns the connectivity layer that everything else sits on. The parallel with what we have seen in gigafactory builds, where the network is treated as part of the process from day one, is instructive.
How does private 5G change the retrofit arithmetic?
A private 5G network attacks the rollout problem at its two weakest points: cost per point and reliability at density.
On cost, the model inverts. Instead of a cable run per sensor, the plant deploys a small number of indoor radios, typically a handful for a large factory hall, operating in Ofcom's shared access spectrum, licensed to the site and therefore free of the interference chaos of unlicensed bands. Battery-powered wireless sensors then attach in minutes: magnet-mount a vibration sensor on a bearing housing, commission it from a tablet, walk to the next one. The marginal cost of the thousandth monitored point is the sensor plus minutes of labour, which is the arithmetic a plant-wide rollout actually needs. The awkward pumps in the basement, the fans on the roof, the compressor house across the yard, points that would never have justified a cable, come into scope.
On reliability, cellular's characteristics fit the application. Licensed spectrum means the vibration waveform from the critical gearbox is not contending with a forklift scanner's retries; the network can prioritise classes of traffic, so a machine alarm always beats a firmware download; and coverage is engineered once for the volume of the building rather than patched access point by access point. High-frequency waveform capture, the difference between "vibration is up" and "outer-race bearing defect, order the part", needs meaningful uplink bandwidth from points all over the plant, and this is precisely the uplink-heavy profile private 5G handles well and WiFi handles grudgingly.
The honest trade-offs: a private network is a bigger initial commitment than a WiFi extension, and for a small plant with a few hundred points and a benign radio environment, WiFi or proprietary sensor mesh may genuinely be enough. The case strengthens with scale, with metal, with existing WiFi congestion, and above all with the number of other applications, AGVs, wash-down-proof connectivity, quality cameras, tablets, that will share the same network. Condition monitoring rarely justifies the network alone; it is usually the second or third application that makes the stack unarguable.
What does good look like in a UK plant?
A realistic plant-wide programme looks like this. Start from criticality, not coverage: rank assets by the cost of their failure, and let the top slice, typically the 10-20% of machines behind 80% of downtime pain, define phase one. Specify sensors against failure modes (vibration for rotating equipment, thermal for electrical, ultrasonic for air and steam leaks) rather than buying one of everything. Put the network in as infrastructure with an owner and a service level, sized for the full estate even if phase one only lights up part of it; resurveying and re-architecting per phase is where costs quietly double. And wire the outputs into the maintenance planning system people already use, because a dashboard nobody actions is just a more expensive way to be surprised.
UK plants do not need to look far for evidence this works: the Manufacturing Technology Centre in Coventry and the National Manufacturing Institute Scotland both run private 5G testbeds where exactly these retrofit patterns are demonstrated, and the West Midlands trials at Worcester Bosch showed real-time machine monitoring over 5G on a working production line several years ago. The technology has been derisked; what remains is the deployment discipline.
Where should a plant start?
With two numbers and a walk. The numbers: your true cost per hour of unplanned downtime on your most critical line, and your current split of planned versus unplanned maintenance hours. The walk: around the plant with your maintenance lead, listing the assets whose failure would hurt most and noting how many of them have any instrumentation at all. Those three artefacts, cost, split, list, are the entire skeleton of the business case, and they usually make the decision obvious.
We build and manage private 5G networks for manufacturers who need the factory floor to go wireless without a telecoms department to run it, sized and priced for mid-sized UK plants rather than multinational flagships. If your condition monitoring programme is stuck at the pilot, talk to us about the network that gets it to the whole plant.