Making thickness data from every mine site mean the same thing

Why a mining CML is not a refinery CML

Thickness monitoring practice was written for process plants where a circuit corrodes at a broadly consistent rate, so a modest number of well-chosen fixed points can represent the whole of it. That assumption is what makes a CML useful: the point stands in for its neighbourhood. In slurry, tailings and concentrate service the assumption does not hold. Erosion loss is a function of local velocity, particle size, angle of impingement and solids concentration, and all four change sharply within a single fitting.

The practical consequence is that wall loss in a slurry line is not a rate applied to a surface, it is a shape. The invert of a horizontal run thins while the crown stays near nominal. The extrados of a bend thins in a comma-shaped patch offset from the centreline of the bend. A single point at six o'clock will either sit in the middle of that patch or miss it entirely, and the two outcomes differ by years of remaining life.

So the first thing a group registry has to accept is that the atomic unit in this industry is often a grid or a scan, not a point. That has consequences all the way through: what gets stored, what gets trended, and above all what number is reported upward. Systems built on the one-point-one-value model will force the site to pick a single number, and the choice of which one is exactly where the group's comparability is lost.

Rotating a spool destroys clock-position identity

Rotating slurry spools on a schedule is standard wear management. Turning a spool a third of a turn moves an unworn section of wall into the wear position and buys another campaign out of the same piece of pipe. It is good practice and it is entirely incompatible with a CML defined by clock position unless someone records the event.

When rotation is unlogged, the data does something distinctive. A point that has been steadily thinning returns a reading substantially thicker than the previous one. Nobody believes it, so it is treated as a bad reading, repeated, confirmed, and then either discarded or absorbed into a smoothed trend. Either way the series is now a blend of two different metal locations and the computed rate is fiction. The tell is a population of points across a site that all step upward on roughly the same date, which is worth searching for in any legacy dataset before it is trusted.

Handling it properly is not complicated but it must be explicit. Rotation is an event with a date and an angle. On that event, the existing clock-referenced points are closed with their final readings retained, and a new set is opened against the new orientation. The pipe keeps its identity and its full history; the trend does not pretend that a rotated wall is the same wall.

You cannot average a minimum

The most damaging aggregation error in a multi-site group is also the least visible. Site A records the minimum of a twelve-point grid at each wear location. Site B records a single reading. Site C records the mean of a corrosion-mapping scan. Group reporting takes all three, calls them thickness, and produces an average, a distribution and a fleet remaining-life estimate. Every one of those outputs is arithmetic performed on quantities that are not the same kind of thing.

The minimum of a sample is a statistic of the sampling effort as much as of the object. Increase the number of points and the minimum can only go down. This produces a perverse signal at group level: the site that invested in corrosion mapping now reports thinner walls and higher loss rates than the site still taking one reading with a hand gauge, and if remedial spend is allocated on that basis, the group rewards the worse programme. Anyone who has watched an integrity budget get reallocated on the strength of a league table will recognise the outcome.

The fix is definitional rather than technical. Store the measurand explicitly, store the grid definition and point count alongside the value, and refuse to compute a cross-site comparison between readings whose measurands differ. Where a comparison is genuinely needed, it has to be built from the underlying grid, not from the summarised number, which is another argument for storing the whole grid rather than the one figure a site chose to report.

No regulator is writing your standard for you

Operators arriving in mining from oil and gas often expect a prescriptive external regime and find that there is not one for the process plant. MSHA 30 CFR Parts 56 and 57 address mine safety and health without setting a thickness inspection programme. Pressure vessels and boilers on site fall under jurisdictional regulation — provincial registration and CSA B51 in Canada, state authorities in the United States, AS/NZS 3788 in Australia and New Zealand — but that covers a subset of the equipment and says little about slurry conveyance.

Tailings governance has moved considerably faster than plant integrity governance. The Global Industry Standard on Tailings Management raised board-level expectations about the whole conveyance chain, which is often the reason a group suddenly needs to answer a question about the condition of its tailings lines across every operation. That question is asked at group level and can only be answered from group data, which is usually the moment the incomparability becomes a boardroom problem rather than an engineering annoyance.

Absent an external prescription, the corporate standard is the only standard, and a standard that exists as a controlled document but not as a system constraint is a standard that sites follow to varying degrees. Making it enforceable in the registry — required fields, controlled vocabularies, a template that a site inherits rather than reinvents — is the difference between a policy and a practice.

Units, precision and the two-decimal problem

A group with operations in Nevada, Chile, Quebec and Western Australia will have data in inches to three decimals, millimetres to two, and mils as integers. Conversion is trivial; the damage is done by rounding at the point of conversion. A rate computed from readings that were rounded to two decimals of a millimetre after conversion from thousandths of an inch carries a quantisation error that can be comparable to the annual loss on a slow-wearing line.

The rule is to store what was measured, in the unit and precision it was measured in, and to convert only for display. That sounds obvious and is violated constantly, usually by a well-meaning import script that normalised everything to a group standard unit on the way in. Once that has happened the original precision is gone and cannot be recovered from the converted values.

Retirement criteria vary just as widely and matter more. One site works to a code minimum wall, another to a percentage of nominal, a third to a number an engineer selected in 2011 for reasons nobody has written down. Remaining life computed against three different finish lines cannot be compared, and the site with the most conservative criterion will always appear to be in the worst condition. Recording the retirement basis, its justification and its version against each line class is what makes the comparison legitimate.

Lined pipe: define the measurand before you define the point

Rubber-lined and polyethylene-lined pipe is common in abrasive service, and it turns thickness measurement into a question that has to be answered before the technician arrives. Depending on instrument setup, couplant and which interface is gated, a reading on a lined line can represent the liner, the steel substrate, or the sum. Two technicians with different settings will produce numbers that differ by the liner thickness and both will look reasonable in isolation.

Liner and substrate also fail differently and on different timescales. Liner wear is the operational concern and is often assessed by a different method entirely; substrate loss is the pressure boundary concern and only begins meaningfully once the liner is breached. Trending them in one column produces a series that steps and reverses in ways that no corrosion model explains.

The registry has to carry material of construction and measurand on the CML itself, with separate baselines for liner and substrate where both are monitored. This is the same discipline that keeps a titanium autoclave component from inheriting a carbon steel retirement thickness, and it is worth enforcing at the group template level rather than leaving to each site's convention.

The register has to outlive the contractor

Mine site inspection is contracted work, typically retendered every two to three years, and the incumbent usually supplies its own data system. That arrangement is comfortable until the tender is lost, at which point the point identities, the baselines, the grid definitions and the historical readings live inside a system belonging to a company that no longer works for you. What transfers is a set of exported reports, which is to say a set of numbers stripped of the definitions that made them meaningful.

Groups that have been through this twice usually arrive at the same conclusion: the operator owns the register and the asset hierarchy, the contractor holds a role within it and writes readings into it, and the contract specifies data handover in the operator's structure rather than as an export. That reframes the transition. Changing contractor becomes a change of crew and login, not a data migration project performed under time pressure by people leaving the site.

It also changes the standardisation problem. When every site's contractor supplies the system, the group's template can never be more than advisory. When the operator supplies the system, the template is a configuration that every site inherits, and deviations become visible exceptions rather than invisible local habits.

Evaluating a system for group rollout

Test the aggregation, not the data entry. Load real data from two sites with genuinely different practices — one single-point, one corrosion-mapped — and ask the system for a combined remaining-life view. A system that produces a confident answer without flagging the incompatibility is telling you it will let your group make that mistake at scale. The right behaviour is to decline, or to segregate, and to say why.

Then test governance. Can a corporate CML template be defined once and inherited by every site? Can a site record a justified deviation with an approver, and does group reporting show which figures were produced under the exception? Can the standard be versioned, so that tightening the grid density next year does not retrospectively invalidate three years of readings taken correctly under the old rule? Those three capabilities decide whether standardisation is achievable or whether you are buying a shared spreadsheet.

Finally, test the rotation and lining cases with your own spool records, because they are where generic asset software fails quietly. Atlantis builds this registry on Odoo, which means the CML sits in the same system as the work order, the contractor, the equipment calibration record and the technician's certification status, and the group template is configuration rather than a rebuild. Affordable, accessible, fully customisable. Request a demonstration or a scoped quote at info@atlantisndt.com.

Request a consultation