Every thickness program eventually argues about condition monitoring location count. One camp wants more CMLs because coverage feels like safety. The other wants fewer because every location is a recurring survey cost. Both arguments miss the actual standard the program must meet: CML density is defensible when it matches the damage the equipment can experience, and the reasoning is written down.
Placement is a mechanism question
A CML earns its place by answering a question about a credible damage mechanism. That produces placement logic, not a number:
- General, uniform thinning is forgiving. A modest set of representative locations per circuit or vessel section characterizes the rate, because any location tells you roughly what every location would.
- Localized thinning is unforgiving. Erosion at elbows and tees, thinning downstream of injection points, corrosion at mixing points and dead legs: these concentrate damage in predictable places, and the CMLs must sit at those places, not at whatever spot was easy to reach from the ladder.
- Cracking mechanisms are not a CML question at all. No density of thickness readings finds a crack, and a file full of healthy CMLs on a cracking-susceptible circuit is false comfort.
So the honest answer to "how many is enough" is: enough that every credible thinning mechanism has locations positioned where that mechanism does its damage, on every circuit where it is credible. For a clean utility line that may be a handful. For an injection point circuit it is a deliberate pattern around the injection region.
Density should move with evidence
CML count is not fixed at commissioning. It should respond to data in both directions.
Increase density when readings surprise you: rates well above expectation, large scatter between nearby locations, or a mechanism review that adds a new credible threat. Scatter especially is a signal that the current locations are not capturing the distribution, which is exactly when averages become misleading.
Decrease density when years of readings show consistent, low, well-understood rates across a circuit. Pruning is legitimate and often overdue. What makes it defensible is the record: which locations were retired, on what evidence, by whose review. What makes it indefensible is silence, locations that simply stop appearing in surveys with no decision behind the disappearance.
Identity is the part programs actually get wrong
The engineering of CML selection fails less often than the bookkeeping. A CML is only useful if the same physical point is measured every survey, because a corrosion rate is a difference between readings at one place. The failure modes are mundane: a location renamed between database migrations, a drawing that disagrees with the field marking, a technician gauging near the marked point rather than at it, two systems holding two versions of the CML list.
Every one of these breaks the pairing between readings, and broken pairing produces phantom rates: metal that appears to grow, or thins at twice the true rate. Programs then burn effort investigating noise their own records created.
The defensible file
For each circuit, the file should carry the mechanism basis for placement, the CML list with locations tied to drawings, the reading history per CML, and the review trail for any density change. With that in hand, "how many locations are enough" stops being an opinion and becomes a documented consequence of the damage assessment.
intelliSPEC anchors each reading to a fixed CML identity at capture time, so the pairing that rates depend on cannot drift. Bring a scattered circuit history to a demo and see what stable identity does to it.
