Setting the Next Interval from Two CMLs on a Wet Separator
A horizontal carbon-steel separator (SA-516-70, 10 ft ID) has been in wet hydrocarbon service since 2004. CML-1 is a vertical strip at the liquid line; CML-2 is at the bottom of the shell where a water boot was removed years ago. Readings:
- CML-1: 0.560 in (2004 baseline), 0.494 in (2014), 0.458 in (2024).
- CML-2: 0.560 in (2004), 0.470 in (2014), 0.418 in (2024).
Required thickness at both locations is 0.400 in per the original design calculation, with no corrosion allowance remaining. The vessel is being inspected now. Operations wants to run 10 years to the next major turnaround. There is no history of cracking or high-temperature damage; service temperature is 140 °F.
1. Calculate the corrosion rate and remaining life at each CML and state the maximum internal inspection interval available today. Which CML governs?
Read the worked answer
CML-1
- Long-term rate: (0.560 − 0.458) / 20 yr = 0.102 / 20 = 0.0051 in/yr
- Interval 2004–2014: 0.066 / 10 = 0.0066 in/yr; 2014–2024: 0.036 / 10 = 0.0036 in/yr (rate is falling)
- Remaining life at long-term rate: (0.458 − 0.400) / 0.0051 = 11.4 yr
- Half-life: 5.7 yr
CML-2
- Long-term rate: (0.560 − 0.418) / 20 = 0.142 / 20 = 0.0071 in/yr
- 2004–2014: 0.0090 in/yr; 2014–2024: 0.0052 in/yr (also falling, but from a higher base)
- Remaining life: (0.418 − 0.400) / 0.0071 = 2.5 yr
- Half-life: 1.3 yr
Governing interval: CML-2 controls. The next internal inspection is due in about 1.3 years, and the external inspection interval must also satisfy half-life or 5 years, whichever is less - so 1.3 years applies to both in practice.
Judgment points: use the long-term (most conservative) rate when the recent rate is lower, because a single good interval does not prove the mechanism has stopped; a downward trend only earns a lower rate after enough intervals of evidence. And the Table 1 cap of 10 years never rescues you - half-life bites first. Running 10 years to turnaround is not permitted on this data.
2. What legitimate options does the owner have to reach the 10-year turnaround, and what is not an option?
Read the worked answer
Not an option: simply setting the interval to 10 years because no leak has occurred, or averaging the two CMLs, or using the recent (lower) rate only. The interval is the lesser of half the remaining life or the Table 1 cap; you cannot extend past half-life by decree.
Legitimate paths:
- Fitness-for-service per API 579. The remaining life is driven by a 0.018 in margin over required thickness at CML-2. An FFS assessment of the actual metal loss (Part 4 general/local, Part 5 if there is any local thinning) can establish a lower effective required thickness, and with it a longer remaining life, if the assessment supports it. This is the main route to a 10-year run and must be documented and accepted by the inspector (and jurisdiction where required).
- Rerate downward. Reducing MAWP lowers required thickness, extends remaining life arithmetically, and buys interval. It requires engineering, new calculations, a new nameplate, and jurisdiction involvement - a real cost, but honest.
- Arrest the mechanism. Internal coating/liner or a corrosion inhibitor program does not retroactively extend today's interval, but once the rate is re-baselined with post-mitigation readings, future remaining life improves. Pair it with shorter interim inspections until the new rate is demonstrated.
- Shorten, don't stretch. Accept the current limit and plan an internal inspection at ~1.3 years, then reset. Combine with an RBI program (API 580/581) where the jurisdiction allows, but RBI manages risk - it does not make a 1.3-year half-life into a 10-year one without an FFS or rerate basis.
Whichever route is chosen, the FFS or rerate basis, the recalculated rates, and the revised due date all go in the vessel file.
3. A dense UT grid around CML-2 returns one reading of 0.372 in over roughly a 2 in × 3 in area; surrounding readings are 0.415–0.425 in. What does the inspector do?
Read the worked answer
Stop treating it as a general-thinning problem. One isolated 0.372 in reading is below the 0.400 in required thickness, while the surrounding metal is not. That is localized loss (a pit or a corroded patch), and the design formula does not pass at that point.
Actions:
- Map the feature - tighten the grid to establish the actual area and depth, and confirm the reading is real (repeat the measurement, check for an internal or external protrusion affecting the gauge, verify couplant and technique, and consider whether a weld or support lies under the spot).
- Do not accept it by averaging. Averaging neighboring readings to get above 0.400 in would hide a through-thickness failure path. Local metal loss is evaluated with API 579 (Part 4 for general, Part 5 for local) using the measured size of the thin region, not the average of the grid.
- Decide on data:
- If FFS shows the region is acceptable at current and projected conditions, document the assessment, set a short re-inspection interval on that spot, and keep monitoring.
- If FFS rejects it, the component must be repaired (weld build-up or patch), rerated to a lower MAWP, or retired.
- Timing: this is a below-required-thickness finding, so it is acted on now - it does not wait for the 1.3-year interval. If the assessment cannot be completed promptly, reduce severity or take the vessel out of service.
The trap in this question is treating a single bad reading as either (a) noise to ignore, or (b) a reason to average. It is neither: it is a localized damage feature requiring its own assessment.