Piping inspection where the data quality decides the answer

What piping circuit inspection establishes

A piping inspection program has to produce, for every circuit in the unit, a defensible corrosion rate, a remaining life for the limiting component, a next inspection date, and a repair list that someone owns. Nothing about that is exotic. What makes piping harder than vessels is scale and identity: a refinery unit holds thousands of measurement points, and each one is worth exactly as much as the certainty that this year's reading was taken where last year's was.

We build the program around that certainty. Circuits are defined before points are chosen, points are numbered and drawn before readings are taken, and readings are paired against a measured baseline rather than a specification. Where a program is inherited rather than built, we audit it first — the arithmetic is quick, the data archaeology is not. Independent review of that layer sits under our ASNT Level III consulting services.

Circuitization: dividing the unit by damage mechanism

API 570 requires piping to be broken down into systems and circuits before an inspection plan can be written. Systems are often defined at process flow diagram level; circuits are usually defined at P&ID level. The organizing principle is the expected damage mechanism, not the line number — sections of piping sharing a damage mechanism, a material and a set of operating conditions belong in the same circuit, because that is what lets one set of readings speak for the whole group.

Get this wrong and everything downstream inherits the error. A circuit that spans a spec break, a temperature change or an injection point averages two corrosion behaviours into one number that describes neither. Dead legs may be combined into a single circuit only when their anticipated damage mechanisms and corrosion rates are alike. Where the population is large enough that circuit-level judgement needs a formal risk basis, risk-based inspection program design puts likelihood and consequence behind the grouping.

CML selection: which points earn a place in the program

CMLs go where the metal actually thins. API 570 directs attention to elbows and tees, reducers, valve discharge points, dead legs, low points, insulated sections, injection points, soil-to-air interfaces, weld heat-affected zones, and immediately upstream and downstream of material changes such as spec breaks. The allocation must consider service-specific damage mechanisms described in API 571 and API 574 — localized corrosion, stress cracking, CUI and high-temperature hydrogen attack are all named as CML conditions, not just wall thickness.

Density follows risk, not geometry. More CMLs go on circuits with higher leak consequence, higher expected or experienced corrosion rates, higher localized-corrosion potential, more fittings, branches and dead legs, higher CUI potential, greater thickness variability, higher short-to-long-term rate ratios, and greater process variability. Fewer go on long, straight, relatively non-corrosive runs. Where CMLs are substantially reduced or eliminated, a corrosion specialist must be consulted and the reasoning recorded.

CML placement and examination points

A CML is an area; an examination point is a spot within it. Where appropriate, thickness measurements should include each of the four quadrants on pipe and fittings, with particular attention to the inside and outside radius of elbows and tees where flow turns and erosion-corrosion accelerates. As a minimum, the thinnest reading — or an average of several readings within the point — is recorded at each examination point.

Finding the minimum is a scanning task, not a spot-reading task. API 570 states that the minimum thickness at a CML can be located by ultrasonic scanning or profile radiography, with electromagnetic techniques used to identify thin areas that are then measured by UT or RT. A single shot in the middle of a 6-inch square that happens to miss a groove is not wrong arithmetic; it is the wrong measurement, and no downstream calculation can recover from it.

CMLs should be marked on inspection isometrics, and the piping itself may be marked so the same locations get measured every cycle. That physical repeatability is what turns a series of numbers into a trend.

A CML is not a random reading

The practical difference between a CML and a random UT reading is identity across time. A CML has a number, a drawing position, a documented baseline, a history and an owner. A random reading has a value. Two randoms taken a year apart on the same spool produce a difference that looks like corrosion and may simply be two different points on a plate with mill tolerance variation, or a probe placed 40 mm away on the far side of a weld cap.

This is why campaign data collected by a crew that did not draw the isometrics so often fails on arrival. The readings are real, the instrument was calibrated, the technician was certified — and the numbers still cannot be paired against history, so the corrosion rate has to be rebuilt from scratch. We check pairing integrity before we accept any inherited data set, which is the same discipline described in what makes an NDT report defensible.

Corrosion rate from paired readings, and which rate governs

API 570 gives two point-to-point rates. The long-term rate is the initial thickness minus the actual thickness, divided by the years between them, where initial means the first measurement at that CML or the thickness at the start of a new corrosion rate environment. The short-term rate is the previous thickness minus the actual thickness, over the years between those two. Remaining life is actual thickness minus required thickness, divided by the governing corrosion rate.

Both rates are calculated and compared to see which produces the shorter remaining life. The authorized inspector, in consultation with a corrosion specialist, selects the rate that best reflects the current process. When the short-term rate changes significantly from the previously identified rate, the inspector should consult a corrosion specialist to determine the cause rather than simply re-baselining. Required thickness is the greater of the pressure design thickness and the structural minimum thickness — not whichever one is convenient.

Statistical analysis is permitted as an alternative to point-to-point, and must be documented and conservative. It is explicitly not applicable to circuits with significant localized, unpredictable corrosion mechanisms, which rules it out for exactly the circuits where the temptation to smooth the data is strongest.

The nominal-thickness trap

API 570 states plainly that installation inspection should document baseline thickness measurements to be used as initial readings for corrosion rate calculations, in lieu of nominal and minimum design thickness data from specifications, data sheets and drawings. That sentence exists because using nominal as the baseline is the most common corruption in piping data, and the most invisible. Pipe is delivered within a mill tolerance, so a nominal figure can overstate or understate the metal that was actually installed.

The consequence compounds. A long-term rate built on a wrong initial thickness is wrong; remaining life divides a correct margin by an incorrect rate; the next inspection date follows the wrong remaining life; and the MAWP calculation, which uses actual thickness minus twice the estimated corrosion loss before the next inspection, inherits it too. If the calculated rate turns out to disagree with observed loss, API 570 requires the rate used for the next period to be adjusted to agree with the actual rate found.

This is a data governance problem more than an NDT problem, which is why we push baselines, tolerances and provenance into a system of record rather than a spreadsheet. Our Odoo-based ERP for NDT companies holds circuit and CML registers with the audit trail attached.

Injection points, dead legs, soil-to-air interfaces and CUI

Injection points get their own circuit and their own interval. The recommended upstream limit is a minimum of 12 in. (300 mm) or three pipe diameters upstream of the injection point, whichever is greater. The downstream limit is the second change in flow direction past the injection point, or 25 ft (7.6 m) beyond the first change in flow direction, whichever is less. CMLs go on appropriate fittings within the circuit and at both its upstream and downstream limits, and the preferred methods are radiography and UT scanning or closely spaced UT grids.

Corrosion under insulation is inspected on externally insulated carbon and low-alloy piping operating between 10 °F (−12 °C) and 350 °F (175 °C), and can be conducted as part of the external inspection. Dead legs, small-bore piping, threaded connections and soil-to-air interfaces each have their own extent rules, because each breaks the assumption that a circuit's average behaviour describes its worst point. Where these circuits sit inside an OSHA PSM-covered process, the documentation rules in OSHA PSM mechanical integrity NDT apply to every one of these examinations.

What the piping record has to show

API 570 specifies what to record per circuit: material of construction and piping specification, diameter, operating and design pressures and temperatures, flange rating, process fluids, class where RBI is not used, insulation, heat tracing and PWHT status, whether the circuit is a dead leg, injection point or intermittent service, the corrosion rate and remaining life of at least the limiting examination point, the maximum interval for external inspection, the maximum interval for thickness measurement, any unusual localized corrosion mode needing specialized techniques, and features that could corrode rapidly during an upset.

Inspection isometrics carry all significant components, material and spec breaks, diameter, insulation status, secondary piping, all CMLs with enough information to locate them, orientation and legible detail, circuit numbers, continuation drawings, and pipe support locations. Repair recommendations must be tracked with corrective action and date, priority or target date, the circuit identifier affected, and a list of temporary repairs needing follow-up. A management system for reviewing outstanding recommendations is required, not optional.

That register is an asset. Once circuits, CMLs and histories are structured, they can be surfaced against the plant model so an engineer sees the thinning elbow in place rather than in a row of a spreadsheet — which is what our digital twins platform is built to do. Send us your isometrics and last campaign data through contact and we will scope the work: affordable, accessible, fully customizable, quote on request.

Request a consultation