On the factory floor, AI's current role is to flag that a machine is failing while there is still time to act on it, before production stops. How much time that warning allows for comes down to the data reaching the software.
These systems need a reading from every machine they are supervising, arriving continuously and arriving on time. A lot of these projects stall, however, because most factory networks were not built to carry traffic of that kind. When the data stops for a while, anything that goes wrong in that window goes unrecorded, so no warning is produced at all.
Private 5G is what an increasing number of plants are building to close that gap, which is a cellular network built for a single site, on spectrum licensed for use at that location only.
In this post, we cover what AI on the factory floor requires from a network, why an industrial structure makes data transmission challenging, what private 5G can offer and the equipment problems sitting underneath.
How AI can support production on the factory floor
AI gives engineers a faster read on equipment data than they would manage alone, and it picks up faults that would have gone undetected by eye. A good deal of that work happens away from the floor, in scheduling, energy use and quality analytics. On the floor itself, three of the most prominent AI uses are:
1. Failure prediction
Sensors on motors, pumps and gearboxes read vibration and temperature continuously, and an AI model compares those readings against how the machine ran when it was fully operational, so a worn bearing shows up in the data before the machine stops and the repair goes into a planned shutdown. That depends on readings every few seconds from hundreds of points, sustained through continuous operation.
2. Inspection
Cameras photograph each part as it moves past them, and a vision model checks the image against what the part should look like, rejecting anything that does not match. Because the line keeps moving, that check has to be finished in the time it takes the part to travel from the camera to the point where it can be pushed off the line.
3. Fleet management
Fleet software drives the AGVs that carry materials around a site. It tracks where each vehicle is, tells it where to go next and keeps them all on route, which only works while every vehicle stays connected to it. An AGV choosing its own route cannot be wired in, so a vehicle that drops the connection stops until it is restored.
All three depend on machines reporting on themselves without interruption, which rests on the software and, just as much, on the network carrying the data to it.
What makes a factory floor hard to connect?
Some of the challenges in factory floor connectivity come down to the two factors most sites have to work with: cable does not reach equipment that moves freely around a site, and Wi-Fi slows down as more devices join it.
Cable does remain the most dependable option wherever it can be run. Fixed equipment can be wired in without difficulty, and so can anything travelling a fixed path, where a drag chain or festoon system carries the cable along with the machine. What cable cannot do, though, is connect the mobile parts of a factory floor, leaving AGVs and mobile robots to be connected another way.
Naturally, Wi-Fi is what most sites use to fill that gap. However, devices share the available airtime and take turns to transmit, so each one waits longer as more of them connect to the network, and response times stretch at the point the floor is busiest.
Coverage is also divided between access points, and a vehicle crossing the site hands over from one to the next as it goes, which is where connectivity may drop. Industrial buildings make poor conditions for a wireless signal, which might be getting blocked and reflected by metal structures, racking and machinery in motion.
A quality connection on a factory floor, and one that allows AI to operate as it should, means equipment is reached while it is moving, timing is held with hundreds of devices connected at once, and the whole site is covered instead of pockets of the floor.
What is private 5G in manufacturing?
Private 5G is a cellular network built for a single site, using the same technology as the public mobile networks but with its capacity reserved for the organisation running it. The network operates on its own dedicated spectrum, licensed to the site directly or provided through an operator, so nothing outside the plant competes for airtime and timing holds, however busy the floor gets.
A growing number of plants are setting up private 5G networks because the connectivity already installed cannot meet what AI equipment monitoring requires of it. Coverage extends across the whole site, devices stay connected as they move, and the network holds its timing with hundreds of them online at once.
For a factory to run a private 5G network, three elements are needed:
- Licensed spectrum, either allocated to the site directly by the national regulator or supplied through a mobile operator.
- Radio equipment, installed across the site to provide coverage in the same way access points do for Wi-Fi.
- A core network, which manages connections and directs traffic. This can sit on a server inside the plant, keeping data on site and shortening the round trip, or be hosted by the provider as a managed service.
Private 5G rarely replaces Wi-Fi entirely, and most sites run both. Wi-Fi continues to handle general connectivity and handheld devices, while private 5G takes on the more critical tasks, where a delay or a dropped connection would mean operational disruption.
Benefits of using private 5G in manufacturing
Predictable response times
Airtime on a private 5G network is scheduled, so the time a message takes to arrive stays within a known range instead of varying with how busy the network is.
Fleet software depends on this to direct AGVs safely, since a delayed instruction can bring a vehicle to a stop in an aisle or send it into the path of another.
Device density
A single cell supports thousands of connected devices without performance dropping away, which makes AI condition monitoring viable across a whole site rather than on selected machines.
Coverage and mobility
Cells cover large indoor and outdoor areas, including yards and multiple buildings, and devices hand over between them without dropping their connection. Vehicles that would lose contact when crossing between Wi-Fi access points stay connected as they move.
Security and data sovereignty
Every device is authenticated by a SIM issued by the plant, so nothing connects without being recognised first. Traffic can also be kept on site, which is important when production data is not permitted to leave the building.
Wireless retrofitting
Connectivity can be added to existing equipment and reconfigured lines without running new cable through concrete floors or interrupting production. This is how sensors get onto machines that are already installed and running.
The equipment gap underneath
A greater challenge, and the reason more sites have not done this already, lies with the equipment on the other side of the connection, since most of it was installed years ago- in some cases even decades- when continuous monitoring was not a requirement.
Older controllers were built to run a machine and report a fault once it had occurred, which means they report a status and an alarm code. An AI model needs a stream of readings to learn from, and that is a different kind of output altogether.
Drives, sensors and instrumentation of the same generation have the same limitation, in that they are working exactly as they were designed to, and that design never included reporting on themselves.
Replacing those controllers and drives across a site, however, is not realistic, given the capital cost, the requalification involved and the downtime required. Most plants retrofit instead, fitting sensors directly to the machinery already in service and connecting them to the network independently, so the readings reach the AI model without the original control system being involved at all.
That approach only holds while the machinery underneath it keeps running. Equipment of that age eventually needs replacement controllers, drives and modules, and a good deal of it may no longer be in production, so availability becomes the limit on how long a monitored machine stays in service.
If you are keeping older automation equipment running, browse our product range by part number or by manufacturer, or get in touch and we will help you source what you need.
FAQs about private 5G in manufacturing
What are the benefits of using private 5G in manufacturing?
With private 5G in manufacturing, response times stay within a known range, thousands of devices connect to a single cell while performance holds, and coverage reaches the whole site instead of pockets of it.
How is private 5G different from Wi-Fi in a factory?
Wi-Fi devices share airtime, so response times stretch as more of them connect. Private 5G schedules airtime instead, which holds timing steady under load.
What is the difference between private 5G and public 5G?
Public 5G is shared with everyone else on the mobile network, while private 5G serves a single site on dedicated spectrum, so its capacity is reserved for that plant only.
Can private 5G connect to existing PLCs?
Not directly, as PLCs have no cellular connectivity of their own. Sensors are fitted to the machinery independently, so their readings reach the network without the PLC being involved.
Does private 5G work with older machinery?
Yes, since the network connects to sensors on the machine rather than to the original control system.
Does private 5G replace wired connections on a factory floor?
No. Cable remains the most dependable option for fixed equipment, but private 5G covers what cable cannot reach, alongside Wi-Fi handling general connectivity.
