The three lives of a pressure vessel.
A 16-foot horizontal separator on a hot Permian pad starts as a number in a process simulator — and ends, twenty years later, as a torn-up fitness-for-service report on a maintenance engineer’s desk. In between are three distinct lives, three distinct languages, and three completely different definitions of what “safe” actually means. This is how each life is supposed to talk to the next — and what we keep getting wrong.
The vessel lifecycle, annotated.
What the design engineer owns.
Where math becomes steel.
Where the design gets second-guessed.
Where a vessel is just variables.
The first life is mostly arithmetic. A process licensor hands you a heat-and-material balance with two columns: P and T. Maybe P = 150 psig, T = 160°F. The licensor’s job ends there. Yours begins.
The instinct of every junior engineer is to plug those numbers straight into the shell-thickness formula and call it design pressure. That is wrong. Design pressure is not operating pressure. It is operating pressure plus a margin that exists because real processes upset, control loops fail, and PSV setpoints drift. The conservative practice — and the one written into most owner specifications — is the higher of operating + 25 psi or operating × 1.10, with a floor of 50 psi. Design temperature is operating + 50°F, sometimes more if the process has any chance of seeing a fire case.
Once you have P_design and T_design, the rest of Section VIII Division 1 is mostly a long table of formulas indexed by geometry. The one that runs the world is UG-27 (c)(1) for the cylindrical shell under internal pressure on the circumferential stress:
P = design pressure (psi, internal)
R = inside radius (in)
S = maximum allowable stress (psi), Table 1A
E = joint efficiency (0.70 spot RT → 1.00 full RT)
Three things matter about this formula that the textbook usually skips. First, S is not a single number for a single material — it varies with temperature. SA-516-70 is 17,500 psi at room temperature and falls below 14,300 psi by 700°F. If you size at room and operate hot, you’ve under-designed. Second, E is a choice you are making on behalf of the fabricator and the NDE department; pick 1.00 and you commit them to full radiography on every long-seam butt weld. That decision will appear, untranslated, on the shop floor twelve weeks later. Third, the formula gives you the minimum required thickness — not the thickness you order. To that you add corrosion allowance, undertolerance on the plate mill certificate (typically 0.01"), and rounding to the next commercial plate gauge.
The corrosion allowance argument
Every job has a meeting where someone tries to talk the lead engineer out of corrosion allowance. It’s clean sweet gas, why are we adding 1/8 inch? The honest answer is that the process licensor told you what’s in the stream at the design case. Nobody can tell you what’s in the stream at year fifteen, after the field has produced through three different reservoirs, two compression upgrades, and one acid-stimulation campaign that nobody remembered to copy you on.
The numbers most operators settle on:
| Service | CA (inch) | CA (mm) | Basis |
|---|---|---|---|
| Sweet gas, dry | 1/16 | 1.6 | Mechanical wear |
| Sweet gas, wet | 1/8 | 3.2 | CO₂ partial pressure |
| Sour gas (H₂S < 50 ppm) | 1/8 | 3.2 | NACE MR0175 minimum |
| Produced water, oxygenated | 1/4 | 6.4 | O₂-induced pitting |
| Crude separator, 2-phase | 1/8 to 3/16 | 3.2 to 4.8 | Water dropout zone |
The head you didn’t think about
Most engineers spend ninety percent of their design effort on the shell and ten percent on the heads. The vessel manufacturers spend it the other way around. The standard 2:1 semi-ellipsoidal head — the one that costs three times the shell per pound — is governed by UG-32 (d):
Notice that the head is thinner than the shell for the same pressure and material — the geometry distributes stress more efficiently than a flat cylinder ever will. This is also why fabricators love forming heads and hate forming nozzles: the stress concentration around a nozzle in the head is twice what it is in the shell, which is why every nozzle larger than NPS 4 gets a reinforcement pad and the calculation in UG-37 that determines its size. Reinforcement is not a styling choice. It is an area-replacement calculation: the metal removed by the nozzle bore has to be replaced, within an envelope around the opening, by metal that wouldn’t otherwise be carrying load.
Where it stops being math and starts being steel.
The U-stamp shop receives your data sheet and immediately starts asking questions the data sheet didn’t anticipate. Which plate mill is supplying SA-516-70 this quarter? Are we rolling the shell course longitudinally or circumferentially? What is the seam orientation relative to the head-to-shell weld? Each answer changes the inspection plan — and the cost — by a measurable amount.
The standard fabrication sequence for a horizontal separator looks roughly like this:
E = 1.0) commits this stage: every butt weld gets 100% radiography. Nozzles get magnetic-particle or liquid-penetrant inspection. Findings are mapped on a weld map and signed off by an ASNT Level II inspector.Why post-weld heat treatment isn’t optional
Welding pumps two things into the joint that you didn’t ask for: residual tensile stress and a heat-affected zone (HAZ) with grain structure unlike the parent metal. For carbon steel above about 1¼” nominal thickness — or for any thickness in sour service per NACE MR0175 — Section VIII makes post-weld heat treatment (PWHT) mandatory. PWHT raises the whole vessel to roughly 1,150°F, holds it for an hour per inch of thickness, then cools at a controlled rate. The residual stress relaxes. The HAZ tempers. The vessel that comes out is metallurgically a different object than the one that went in.
Operators sometimes try to escape PWHT because it adds two weeks and tens of thousands of rupees of furnace time to the schedule. The trade is usually not worth it. A vessel installed in sour service without PWHT is a candidate for sulfide stress cracking (SSC) within months, and the repair cost on a cracked vessel — including the inevitable production loss — dwarfs any PWHT savings. The right place to fight the PWHT cost is upstream, in the design stage, by selecting a material grade and thickness that doesn’t need it.
The hydrotest
The final shop-floor act is the hydrostatic pressure test, UG-99, at 1.30 × MAWP corrected for the difference between test-temperature stress and design-temperature stress. The vessel is filled with water — never gas, never pneumatically unless you have very specific written justification — pressurized, held for a minimum of 30 minutes, and inspected for leaks at every weld and nozzle. A vessel that leaks at hydrotest leaks back to the welding station; one that passes gets a U-stamp riveted to the nameplate.
Where everything you designed gets second-guessed.
The vessel arrives at site, is set on its saddles, piped up, insulated, and goes into service. From that day forward it is governed by a completely different document: API 510, Pressure Vessel Inspection Code. API 510 doesn’t care what your original design calculation said. It cares about what the wall thickness is now, what the corrosion rate has been, and how much remaining life that gives you before the vessel falls below its minimum required thickness, t_min — which is the UG-27 number, recomputed with whatever has changed about the process.
The inspection workflow is built around three numbers:
t_initial = thickness at commissioning, from MDR
t_actual = current minimum, from UT scan grid
t_min = code-required thickness for current MAWP
The UT scan grid is where field engineering meets statistics. The inspector lays out a grid of measurement points — typically 6 inches on a side for general corrosion, tightened to 1 inch in known erosion zones like inlet impingement or below the normal liquid level — and pings each one with an ultrasonic thickness gauge. The minimum reading at each grid point becomes t_actual for that location. The next-inspection interval is governed by API 510 §6: half the remaining life, with a hard ceiling of 10 years and a floor of 5 years internal / 10 years external for most services.
Fitness-for-service: when the math runs out
Eventually every vessel meets a measurement it doesn’t pass. A 0.480" thickness reading shows up in a vessel that’s supposed to be 0.500". The corrosion rate over the last five years suddenly accelerated. The PSV setpoint creeps. When this happens, the vessel doesn’t automatically come out of service — it goes into a Fitness-For-Service (FFS) assessment per API 579-1 / ASME FFS-1. FFS is a tiered evaluation:
- Level 1 — a screening calculation, often done in a spreadsheet. Does the remaining wall, at design pressure and temperature, give an acceptable margin against UG-27?
- Level 2 — a more detailed evaluation that accounts for local thin areas (LTAs), groove-like flaws, and bulging. Requires a qualified pressure vessel engineer.
- Level 3 — full finite-element analysis, often with measured material properties from coupons cut from the vessel itself. Expensive, slow, but defensible.
An FFS that comes back unfavorable doesn’t always mean retirement. It can mean derating — re-stamping the vessel at a lower MAWP that matches the corroded wall — or installing weld-overlay repairs followed by a re-hydrotest at the reduced pressure. Both options preserve the asset; both close the loop back to ASME VIII for the repair calculation. The U-stamp the vessel left the shop with twenty years ago is permanent, but the operating envelope behind it is a living number.
The handover document nobody reads
What ties the three lives together is the Manufacturer’s Data Report (U-1A or U-1). It carries the original design pressure, temperature, material, joint efficiency, corrosion allowance, NDE results, hydrotest record, and U-stamp. It is the only document the inspector twenty years later will have that tells them what the vessel actually was on its first day. Most jurisdictions require it to be filed with the National Board of Boiler and Pressure Vessel Inspectors; most operators file it in a project archive that nobody opens again.
The one habit that distinguishes a strong equipment engineering team from an average one isn’t the fancy FEA software or the corrosion-allowance argument. It is this: they write the inspection plan on the same day they finalize the data sheet. They put the future scan grid, the predicted high-wear locations, and the expected corrosion rate on a page that becomes Appendix A of the MDR. When that vessel shows up at its first internal inspection eight years later, the API 510 inspector doesn’t have to guess where to scan. The design engineer told them — eight years ago, in writing.
That habit is what keeps the three lives talking to each other. Everything else in this article is downstream of it.
The shell-thickness math, automated.
The Design Hub Premium 3-Phase Separator Sizing calculator runs the full UG-27 / UG-32 / UG-37 stack — including reinforcement-pad area, ρV² nozzle sizing, and L/D optimization — in your browser. Built on the same API 12J / GPSA / ASME VIII practice this article uses.
