In short: Corrosion under insulation (CUI) eats insulated pipework and vessels out of sight, and the traditional way of finding it, stripping cladding off from a scaffold, is slow and expensive. Wireless thickness and moisture sensors can watch the worst lines continuously, but only if the site has a network that reaches every pipe rack and carries the data back to the integrity team.
Key Takeaways
- CUI hides where inspectors cannot look — moisture trapped under cladding corrodes carbon steel fastest between roughly -4°C and 175°C, the operating range of a large share of refinery and terminal pipework.
- Sensors replace the scaffold for the worst lines — permanently fitted ultrasonic thickness and moisture sensors report every day, so inspectors strip insulation where the data points rather than on a fixed sample.
- The network is the gating factor — hundreds of battery sensors across steel pipe racks need gateways and a site-wide backhaul, and a private 5G network gives one managed layer for sensors, inspectors and drones.
In a nutshell

What is corrosion under insulation, and why is it so hard to find?
Corrosion under insulation is external corrosion that forms on pipes, vessels and tanks beneath their thermal insulation and weather cladding. Water gets in through damaged jacketing, poorly sealed penetrations and failed caulking at supports, soaks into the insulation and then sits against the steel, often warm, often laced with chlorides leached from the insulation itself. The guidance the industry works from, AMPP's SP0198 standard (formerly NACE) and API's RP 583, puts the danger zone for carbon steel at roughly -4°C to 175°C, with austenitic stainless steels prone to external chloride stress corrosion cracking across much of the same range.
The difficulty is that the damage is invisible by design. The insulation that keeps the process at temperature also hides the pipe wall from every visual inspection, and a cladding sheet can look perfect while the steel underneath has lost half its thickness. Lines that cycle in and out of the danger zone, such as intermittent steam tracing, relief headers and lines that sit idle between campaigns, are particularly prone to it, because they wet and dry repeatedly.
The consequences are well understood. ExxonMobil's long-quoted estimate is that CUI accounts for 40 to 60 per cent of piping maintenance costs, and on a COMAH site a through-wall failure on a hydrocarbon line is a loss-of-containment event, with all the regulatory scrutiny that follows.
Why does CUI keep catching UK operators out?
Firstly, the plant is old. Much of the UK's refining and terminal infrastructure dates from the 1950s to the 1970s: Fawley, Stanlow, Pembroke and the Humber refinery, along with gas terminals such as St Fergus and Bacton, and Sullom Voe in Shetland. HSE's Key Programme 4 (KP4) on ageing and life extension, which ran offshore from 2011 to 2013, pressed duty holders to show they understood the condition of their ageing plant, and CUI was among the degradation mechanisms in scope. The same mechanisms apply onshore, and the Pressure Systems Safety Regulations 2000 require a written scheme of examination that has to account for them.
Secondly, the inspection model is expensive per metre. To check a section of insulated line, a contractor typically erects scaffold, strips the cladding and insulation, cleans and inspects the steel, then reinstates the lot. On elevated pipe racks the scaffold and reinstatement can easily cost more than the inspection itself. Operators therefore sample, using risk-based inspection (RBI) to pick the locations most likely to be corroding, and accept that the rest of the line goes uninspected until the next cycle.
Thirdly, turnaround windows are tight. Refining margins in the UK have been squeezed hard enough that Grangemouth stopped refining in 2025 and Lindsey closed the same year. The sites that remain run longer between shutdowns and want every turnaround day spent on work that the data says is needed. A CUI programme built on sampling sits awkwardly with that, because the sample is chosen on probability rather than evidence.
How does continuous corrosion under insulation monitoring work?
Continuous CUI monitoring puts permanent sensors on or under the insulation at the locations RBI ranks highest, and reports their readings every day without anyone climbing the rack. There are three main families of sensor, and most programmes use more than one.
Wall thickness sensors. Permanently mounted ultrasonic transducers measure remaining wall thickness at a fixed point, typically once or twice a day, on batteries rated for several years. Emerson's Permasense sensors, which grew out of a research programme between BP and Imperial College London, are the best-known example and report over WirelessHART. Because the sensor never moves, it can resolve a change of a fraction of a millimetre over a few months, which a manual spot reading from a different contractor on a different day cannot. Bristol-based Inductosense takes a different approach, embedding passive thickness sensors under the insulation and reading them with a handheld probe, which removes the strip-and-reinstate step but still needs a person on the rack.
Moisture and corrosion-condition sensors. Since water is the precondition for CUI, sensing water directly gives an earlier warning than waiting for metal loss. Cambridge spin-out CorrosionRADAR, for example, runs a long sensing cable beneath the insulation that detects moisture and corrosion along its length, so one sensor covers tens of metres of line, rather than a single point.
Periodic scanning tools. Pulsed eddy current, digital radiography and guided wave testing can all assess wall loss through insulation without stripping it. These are campaign tools, used by an inspector on the day, and they produce large data files that need to reach the integrity team quickly to be useful.
The value comes from combining the three. Moisture sensors tell the integrity engineer where water has got in; thickness sensors confirm whether the steel is actually thinning; and targeted strip-and-inspect, guided by both, replaces the blanket sample. The RBI model gets updated from measured data instead of assumed corrosion rates.
Why is connectivity the bottleneck for CUI monitoring?
A CUI programme of any useful size means hundreds of sensors spread across a site that may cover a square kilometre or more, much of it a forest of steel pipe racks, vessels and tank bunds. Each sensor sends a few bytes a day, so the hard parts are reach and management.
WirelessHART and ISA100 field networks form meshes, with each gateway handling a limited number of devices within a few hundred metres, and a mesh that is dense in one process unit can be patchy in the next. LoRaWAN gateways reach further but still need a backhaul from each gateway to the control network. On most UK sites that backhaul is either a fibre run that someone has to trench and pull through a hazardous area, or a legacy Wi-Fi bridge that the IT team would rather not own. The result is that sensor deployments stall at the first unit, because extending them to the tank farm or the jetty means another civils job.
The data also has to land in the right place. Thickness and moisture readings belong in the integrity management system next to the RBI model and the inspection history; pulsed eddy current scans and drone footage need to come off the rack while the inspector is still there, so a dubious result can be re-shot without a second scaffold. A site network built for the control system, segregated for good reasons under IEC 62443, is rarely set up to carry either.
Where does private 5G fit, and where does it not?
Private 5G earns its place as the site-wide layer that the sensor gateways, inspectors and inspection tools share. A handful of radios mounted on existing structures, outside the classified zones where possible, can cover a refinery's process units, tank farm and jetties, giving every WirelessHART or LoRaWAN gateway a backhaul without new fibre. The same network carries an inspector's tablet uploading a pulsed eddy current scan from the top of a rack, a drone streaming high-definition video of a flare line, or a crawler robot working along a pipe bridge. Because the operator owns the SIMs and the core, each class of device sits in its own network slice or VLAN, which suits an OT security model that keeps integrity data away from the process control network.
The UK licensing position helps. Ofcom's shared access licences in the 3.8 to 4.2 GHz band give an operator its own spectrum on site for a modest annual fee, so the network does not depend on a public operator's coverage, which at many coastal terminals is thin.
There are honest limits. The sensors themselves will mostly stay on WirelessHART, ISA100 or LoRaWAN, because those protocols are built for devices that run for years on a single battery, and 5G RedCap and NB-IoT sensors with the right hazardous-area certification are only starting to appear. Radios inside Zone 1 or Zone 2 areas need Ex-certified enclosures or must be sited outside the zone, which constrains the radio plan. And a site that already has good fibre to every unit, and a well-run industrial Wi-Fi network, may find that extending what it has costs less. Private 5G makes most sense where the CUI programme is one of several things the site wants to connect, alongside gas detection, lone worker devices and digital permit to work, so the network cost is shared across several budgets.
What does a sensible first CUI monitoring deployment look like?
We would start small and measurable. Pick one process unit or pipe rack that RBI already ranks as high CUI risk, typically a section with intermittent service, steam tracing or a history of findings. Fit 50 to 100 sensors: thickness sensors at the locations the inspection history flags, and moisture sensing along the lines where the cladding is oldest. Put a private 5G radio, or two, where they cover the unit and its route back to the integrity office, and connect the sensor gateways to it.
Then run it for twelve months alongside the existing inspection plan, and compare. The questions to answer are concrete: how many strip-and-inspect locations did the sensor data rule in or rule out, how much scaffold did that avoid, and did any sensor flag wall loss that the sampling plan would have missed? If the numbers stack up, the radio plan already covers the next unit, and extending the programme becomes a sensor purchase rather than a civils project.
Our team at Aerix designs and runs private 5G networks for UK industrial sites, from the radio plan through to the core, and we are happy to look at a site layout and RBI ranking with an integrity team to see whether the sums work. Get in touch through our oil and gas sector page to arrange a site survey.
