Why "annually" is the wrong answer
Fixed intervals are popular because they are easy to schedule, and they are wrong in both directions. A stainless line in benign service gets inspected nine times without a millimetre of change, while a carbon steel line downstream of a control valve loses its wall between two visits.
The alternative is not complicated. It is arithmetic on the readings you already have, and every in-service inspection framework used in Australia — AS/NZS 3788 for pressure equipment, and the API series (510 for vessels, 570 for piping, 653 for tanks) that most sites' written schemes are built on — arrives at the interval the same way.
The four numbers
tactual — the current measured minimum thickness at the location.
trequired — the minimum thickness the equipment actually needs. This is not the nominal wall it was built with. It comes from the design code: the pressure design thickness plus any structural minimum, and it is the number an engineer supplies, not one the inspector invents.
Corrosion rate — millimetres per year, from two readings at the same location:
- Long-term rate = (thickness at commissioning − thickness now) ÷ years in service
- Short-term rate = (thickness at the previous inspection − thickness now) ÷ years between the two
Calculate both and use the higher one, unless there is a documented reason the short-term figure is spurious. A rate that has recently accelerated is the one that matters; averaging it back over twenty years hides it.
Remaining life = (tactual − trequired) ÷ corrosion rate
Then the interval: half the remaining life, or the code maximum, whichever is shorter. The half is not conservatism for its own sake — it guarantees at least one more measurement before the wall reaches its limit, so a rate that has changed is caught with time still on the clock.
A worked example
A carbon steel process line, 8.0 mm nominal wall. The engineering assessment gives a required minimum of 4.5 mm.
| Reading | |
|---|---|
| Baseline, commissioning 2020 | 7.9 mm |
| Inspection 2023 | 7.3 mm |
| Inspection 2026 | 6.4 mm |
Long-term rate = (7.9 − 6.4) ÷ 6 years = 0.25 mm/yr
Short-term rate = (7.3 − 6.4) ÷ 3 years = 0.30 mm/yr
The short-term rate is higher, so it governs. Something changed after 2023 and the line is corroding faster than its history suggests.
Remaining life = (6.4 − 4.5) ÷ 0.30 = 6.3 years
Next inspection = 6.3 ÷ 2 = 3.1 years — call it 3 years, and note in the report that the rate has accelerated so the trend is worth explaining before then.
Had only the long-term rate been used, the answer would have been 3.8 years. The difference between those two numbers is the entire value of taking readings at the same place twice.
The first survey cannot set an interval
A single set of readings gives you today's wall thickness and nothing else. There is no rate, so there is no remaining life and no interval — only a baseline.
Where there is no history, the usual approach is to establish the baseline now, repeat at twelve months, and let the second survey produce the first real interval. In aggressive service, or where the fluid, temperature or velocity has recently changed, six months is more sensible. From the second survey onward the equipment sets its own schedule.
Repeatability is the whole game
A corrosion rate is a difference between two small numbers, so anything that makes the second reading non-comparable with the first destroys it.
- Same locations. Readings are taken at marked condition monitoring locations on a recorded grid. A reading taken 100 mm away from last time on a pitted line can show 0.4 mm of "loss" that is really 0.4 mm of relocation. This is the most common false alarm in thickness monitoring, and the reason our surveys deliver the grid, the photographs and the location references with the numbers.
- Same technique. Dual-element or single-element, echo-to-echo or first-echo, coating included or excluded — a change in any of these shifts the readings systematically.
- Temperature. The velocity of sound in steel falls as the metal heats. Readings taken on a hot line without compensation read thick. If the line is hot at one inspection and cold at the next, the difference is partly instrument, not corrosion.
- Resolution. Reporting to 0.1 mm and then dividing by a twelve-month gap gives a rate with a very large relative error. Longer gaps between the early surveys produce better rates, which is another argument against inspecting a slow-corroding asset too often.
Five things that break the calculation
Localised attack. The whole method assumes general wall loss. Pitting, microbiologically influenced corrosion and under-deposit attack are local, and a grid on 300 mm centres can pass straight over a pit that will perforate first. Where the damage mechanism is localised, the answer is tighter grids or ultrasonic corrosion mapping, not a faster schedule on the same sparse points.
Corrosion under insulation. The readings will be perfectly reassuring everywhere you can reach, and the damage will be under the cladding where you cannot. CUI is worst on intermittently heated lines in the 50–175 °C band, and it is found by removing insulation at targeted locations or by scanning — not by trending accessible points.
Erosion at fittings. Flow-accelerated damage concentrates at elbows, tees, reducers and immediately downstream of control valves and orifice plates. If the CML sits on the straight run, the survey will report a healthy line right up until the bend fails. The location has to be on the outside radius of the bend.
A change in service. New fluid, higher temperature, higher velocity, a different chemical dosing regime, a coating applied or a section replaced — any of these invalidates the historical rate. The trend restarts.
Nobody owning the numbers. A thickness survey that arrives as a PDF and goes into a folder produces no interval at all. The readings have to be held against the asset so the next survey is a comparison rather than a fresh start.
Where the other methods fit
Thickness testing tells you how much wall is left. It does not tell you whether there is a crack in a weld — that is ultrasonic flaw detection or a surface method — and it will not find wet insulation, which infrared thermography picks up across a whole run in minutes and which is usually what is driving the corrosion in the first place. On most of our jobs the thickness grid, a thermal scan of the insulation and magnetic particle on the critical welds are done in one mobilisation, by the same crew, on the same ropes.
If you are weighing up whether a survey is worth doing at all, the cost-of-late-detection calculator will put a number on the difference between finding wall loss on a schedule and finding it on a shutdown.
Frequently asked
Is there a maximum interval regardless of the calculation?
Yes. Codes cap it — commonly five years for piping and ten for pressure vessels — and the site's written scheme of inspection or its regulator may set something shorter. The calculation gives you the interval; the cap is the ceiling on it. Hazard level under AS 4343 also affects what the scheme requires.
What if the corrosion rate comes out as zero?
That is a real and common result on stainless or well-protected carbon steel, and it means the remaining life is not governed by corrosion. The interval then falls back to the code or scheme maximum, and the survey is confirming that nothing has changed — which is worth doing, because service conditions do change.
Can you use one reading per line?
Not usefully. A line has different damage at the bottom of a horizontal run, at bends, at dead legs and at supports. The grid is designed around where the mechanism is expected to act, and one point per line will find the damage only by luck.
How much surface preparation is needed?
For a sound coating, often none — echo-to-echo measurement reads the steel through paint. Heavy scale, rust scabs and blistered coating have to be removed at the reading points, which is worth knowing when scoping, because on a large grid the preparation can take longer than the readings.
Do you have to shut down to take readings?
No. Thickness testing is done from the outside on live equipment. The practical limits are surface temperature and access — and access at height is what rope access is for.