

A DN80 ball valve came out of a urea plant's high pressure carbamate loop after fourteen months and the inspection report opened with a compliment: the bore was in excellent condition. It was. Bright, smooth, no measurable loss of wall anywhere the flow could reach. Then the inspector put a borescope into the cavity behind the upstream seat and found a pit four millimetres deep, a crack running out of that pit into the body wall, and a downstream seat band worn back to a rounded lip.
Same alloy, same fluid, same temperature, same fourteen months. The only difference between the bore and the cavity is that one of them is swept and the other is not. That sentence is the whole of this application, and almost every specification written for it misses the point by treating corrosion resistance as a property of the metal rather than a steady state that has to be maintained. In this loop the metal is not resisting anything on its own. It is being held in a passive state by a chemical that the plant injects on purpose, and the valve is part of the distribution system for that chemical, whether or not the data sheet says so.
What follows is the arithmetic behind that claim: what the fluid actually is in numbers, why a cavity is a different environment from a bore, why the word grade in urea grade means an alloy rather than a certificate, why one litre of this liquid becomes two hundred litres of gas when the pressure is released, and why the valve after the evaporators fails for the opposite reason. It ends with the seven clauses that would have prevented the pit, none of which are in ASME B16.34.
1. Six duties, one alloy family, almost nothing else in common
A urea plant is a short process with a very tall pressure profile. Carbon dioxide arrives at 150 to 160 bar from the last stage of the compressor, carrying 0.25 to 0.6 percent air by volume that nobody would add if corrosion were not the reason. Ammonia arrives from the pump at 160 to 250 bar and 40 to 50 degrees C. They meet in the reactor at 140 to 160 bar and 180 to 190 degrees C with a nitrogen to carbon ratio of 2.8 to 3.2, and 55 to 70 percent of the carbon converts to urea in a single pass; the rest leaves as carbamate and is recycled. Then the whole mixture is let down to 4 to 6 bar, the carbamate is decomposed and driven off, the water is evaporated, and what is left is a 99.7 percent urea melt at 140 degrees C that has to be prilled or granulated before it sets.
A 2,000 tonne per day plant will have several hundred valves in that description, and the ones that matter for this article are the handful that isolate the high pressure loop, the letdown position, and the melt line. They are bought in one package, from one supplier, usually against one material specification, and they could not be more different in what kills them.
Duty | Fluid | Pressure and temperature | What decides the valve |
|---|---|---|---|
CO2 compressor discharge | dry CO2 with 0.25 to 0.6 percent air | 150 to 160 bar, 120 to 130 C | Dry gas, no lubricity, fine clearance, oxygen is the point of the stream |
Ammonia feed | liquid ammonia | 160 to 250 bar, 40 to 50 C | Stress corrosion cracking of any carbon steel in the wetted path |
High pressure loop isolation | carbamate melt with urea and excess ammonia | 140 to 160 bar, 180 to 190 C | Passive film, ferrite content, cavity turnover |
Letdown to the medium pressure section | the same liquid, flashing | 150 bar in, 5 bar out, 185 to 152 C | Two phase velocity, erosion of the trim, 200 fold volume change |
Evaporator outlet | 99.7 percent urea melt | 0.03 to 0.3 bar, 138 to 142 C | 132.7 C freezing point, biuret above 150 C |
Prilling or granulation feed | melt, near atmospheric | 1 to 3 bar, 132 to 135 C | Six kelvin of margin before the line sets solid |
Read the third row and the fifth row together, because they are the two halves of the trap. The high pressure loop punishes a valve for being too cold and stagnant. The melt section punishes it for being too hot. A single jacket setpoint cannot serve both, and a single material specification should not try.
2. The steel is not corrosion resistant here, it is being protected
Put 316L in carbamate with no oxygen present and it does not sit quietly. The corrosion potential falls into the active region, the chromium oxide film is not repaired as fast as it is dissolved, and the surface starts to move. Add oxygen and the potential jumps several hundred millivolts into the passive region and the same alloy becomes, for practical purposes, inert. The difference in metal loss between the two states is not a percentage. It is one order of magnitude at best and two at worst: a well passivated surface runs at 0.01 to 0.05 millimetres a year, an active one at 0.1 to 1.0.
This is why the air injected into the carbon dioxide feed is the most important corrosion control measure on the plant, and why it is also the most fragile. The quantity is small, 0.25 to 0.6 percent by volume, and it is set by two opposing constraints: enough oxygen to keep every wetted surface passive, not so much that the off gas system carries an explosive mixture and the ammonia converter sees inerts it was not designed for. When a plant trims the air to recover capacity, the corrosion rate does not rise in proportion to the trim. It rises when the last surface loses its film, and the last surface to lose it is always the one with the worst access.
There is a second, subtler consequence for a valve buyer. Because the protection is electrochemical, it needs a continuous supply of oxygen at the metal surface, and it needs the surface to be wet. A dry valve in this loop is a valve with no electrolyte and no corrosion, which sounds like good news until the plant is restarted and the first liquid to arrive is oxygen depleted. The dangerous moment in a urea plant is not the shutdown. It is the first hour after a shutdown that was long enough for every dead volume to consume what it had.
3. The cavity is the one place nobody feeds
A DN80 ball valve has a cavity between the seat and the body wall with a volume of roughly 0.3 to 0.4 litres. When the valve is closed, that liquid is connected to the flowing stream only through whatever leaks past the seat, and a metal seat which is passing its acceptance test does not leak much. Take three millilitres a day, which is far below any leakage class limit and is roughly what a tight metal seat does in this service: four hundred millilitres at that rate takes a hundred and thirty days to turn over once. The liquid in that cavity has an age measured in months.
Note what that implies, because it is the opposite of every other rule in valve engineering. The tighter the seat, the longer the cavity sits, and the worse the cavity gets. A valve that leaks a cupful a day is flushing its own dead volume and will outlive the good one next to it.
Now do the oxygen budget. Four hundred millilitres at a density of 1,100 kilograms per cubic metre is 0.44 kilograms of liquid. Carbamate holds oxygen at the tens of parts per million level, so call it 44 milligrams of dissolved oxygen in the cavity. The wetted area inside that cavity is around two hundred square centimetres. At an active corrosion rate of 0.1 millimetres a year, two hundred square centimetres of stainless loses about 16 grams of metal a year, which is 43 milligrams a day, and oxidising iron to its trivalent oxide consumes roughly a third of its own mass in oxygen. The cavity therefore has something between two and three days of oxygen at the rate it is being consumed, and a hundred and thirty days between refills.
That is the mechanism of the pit in the opening paragraph. The film dies in the first week, the cavity runs active for the rest of the year, and because the corrosion is under a stagnant liquid with no flow to sweep the products away, it does not generalise. It localises, it acidifies its own pit, and it goes down four millimetres in fourteen months while the bore stays bright.
Three design responses follow, and they are more useful than any materials upgrade. The first is to make the cavity small or non existent: a seat carried in the ball, or a seat that closes against the body with no annulus behind it, has less dead volume to poison. The second is to connect the cavity to the upstream side so that it is at line pressure and exchanges liquid every time the valve is stroked, which turns a hundred and thirty day turnover into a shift. The third, and the cheapest, is to stroke the valve. A valve that is opened and closed once a shift refreshes its own cavity, and a valve that has been shut since commissioning has a cavity full of liquid that is thirteen months old. Isolation valves are the ones that fail, precisely because they are the ones that do not move.
4. Why urea grade is an alloy, not a certificate
Standard 316L is not one phase. It contains between three and eight percent delta ferrite, deliberately, because ferrite resists solidification cracking in weld metal and improves resistance to chloride stress corrosion in some services. In carbamate, that ferrite is the weakest link. It is a chromium and molybdenum depleted phase sitting next to austenite, it forms a galvanic couple with it, and the attack follows the ferrite stringers into the material rather than across the surface. A forging with banded ferrite will corrode in bands, and a weld with a ferrite network will corrode along the network.
Urea grade 316L is a different specification with the same name. Nickel is raised to about 14 percent, molybdenum to 2.2 or 2.5, carbon is held at 0.02 or below, residuals are tight, and the ferrite content is capped, typically at 0.6 percent and often at 0.3. Weld metal is fully austenitic, which means the welder has traded away the ferrite that prevents hot cracking and has to earn it back with a qualified procedure, low heat input and, on the thicker sections, a solution anneal at 1,040 to 1,080 degrees C followed by a water quench. There is no post weld heat treatment in this family: heating a sensitised austenitic stainless in the 500 to 800 degree band is how you precipitate the chromium carbides that make it corrode at the grain boundaries.
The trap for a buyer is that the mill certificate does not state ferrite. It states chemistry, mechanical properties and a heat number, all of which can be perfectly compliant while the finished forging sits at four percent ferrite because the forging reduction was wrong. Ferrite is a product of thermal history as much as of composition, and the only place to measure it is on the finished part, with a ferrite scope, at the surfaces that will see carbamate. A certificate that says 316L is not evidence of urea grade and never was.
Option | Ferrite or structure | Pitting resistance equivalent | Where it still fails | Relative cost of the wetted part |
|---|---|---|---|---|
Standard 316L | 3 to 8 percent delta ferrite | 24 to 26 | Selective attack along ferrite stringers and welds | 1.0 |
316L urea grade | 0.6 percent maximum, fully austenitic welds | 26 to 28 | Crevices that lose their oxygen supply | 1.0 to 1.2 |
25Cr 22Ni 2Mo austenitic | fully austenitic, high alloy | 33 to 35 | Cost, welding without filler control, availability | 1.6 to 2.0 |
High alloy duplex for urea | austenite and ferrite, stabilised against carbamate | above 40 | Needs the licensor's own welding procedure, not a generic one | 2.4 to 3.0 |
Zirconium 702 | single phase, immune | not applicable | Hydriding if the water or oxygen chemistry goes wrong, and price | 6 to 10 |
Two options that look sensible and are not. Duplex 2205 is a fine alloy for chloride service and a bad one here, because the ferrite that gives it its strength is exactly the phase that carbamate attacks. Titanium, the reflex answer whenever a chloride line is mentioned, is worse: it is not resistant to carbamate at these temperatures and it has no place in the loop. Both get offered as upgrades by suppliers reading the word corrosive and reaching for a data sheet, and both have been installed and removed.
5. Releasing the pressure is a phase change, not a pressure drop
The high pressure loop exists at 140 to 160 bar for a reason that has nothing to do with mechanical design. The dissociation pressure of molten ammonium carbamate at 185 degrees C is in that same range, so the loop pressure is the pressure at which the melt stops trying to boil itself apart. Below it, the carbamate decomposes endothermically into ammonia and carbon dioxide, taking heat out of the liquid as it goes, which slows the reaction and produces a cold, frothy, two phase mixture that behaves nothing like either phase.
The volume arithmetic is worth doing once because it explains the geometry of every letdown station ever built. A litre of melt at 1,150 kilograms per cubic metre is 1.15 kilograms, which is 14.7 moles of ammonium carbamate, and each mole yields three moles of gas, so 44 moles, which is 986 litres at atmospheric pressure. Compress that to 5 bar and it is still 197 litres. One litre of liquid becomes two hundred litres of gas, and it does so in the space of a few milliseconds inside a valve trim.
What that means for the valve is that the letdown position is not an isolation duty and should never be given to an isolation valve. A ball valve throttling at 20 percent open in this service sees a two phase jet at a velocity that no erosion allowance survives for long, and the seat, which was lapped to a light band, is the first thing the jet finds. The correct arrangement is a dedicated letdown device with an expanded outlet, hardened trim and a body designed for the volume, with an isolation ball valve upstream that is either fully open or fully closed and never in between.
The small mercy is that carbamate is water soluble. Deposits that form on a cool down can be washed out with warm water rather than chipped off, which is not true of sulfur, wax or polymer. The wash water has to be demineralised and low in chloride, because the crevice under a seat is exactly where chloride concentrates, and it has to be warm, because a cold wash crystallises the recycle carbamate somewhere else instead of removing it.
6. After the evaporators the problem inverts
The melt that leaves the last evaporator is 99.7 percent urea at about 140 degrees C, and its freezing point is 132.7. That is seven kelvin of margin, in a line that runs through a building, and it is the reason every valve on it is traced or jacketed rather than insulated. But the jacket that keeps it liquid is also the thing that can ruin the product, because urea does not merely freeze below 133 degrees C. Above about 150 it starts to combine with itself, two molecules losing ammonia to form biuret, and biuret is the impurity that decides whether the prills can be sold as fertilizer or have to be blended down.
The rate is a residence time problem, not a temperature alarm. Saturated steam at 6 bar gauge sits at 165 degrees C, which is 25 kelvin above the melt and squarely inside the band where biuret forms; steam at 4 bar gauge sits at 151, which holds the wall above the freezing point and stays out of trouble. A jacketed valve body with a 6 bar jacket and a dead annulus behind the seat is a small biuret reactor that runs continuously for years, and no product analysis will ever point at it because the volume is tiny compared with the evaporator.
Freezing, at least, is benign in the way that salt is benign: urea contracts when it solidifies, from about 1,220 kilograms per cubic metre in the melt to 1,330 as a solid, so a frozen body does not split and a frozen plug does not seal. What it does is grip the ball. Solid urea around a 80 millimetre ball is worth a breakaway torque several times the actuator's, and the failure signature is a sheared key and a stem with a twist in it, reported as a seized valve. The recovery is easier than in a salt system because the melting point is low enough that a warm wash will do it, provided someone thinks of a warm wash before reaching for a bar on the handwheel.
7. The seat has a thermal ceiling that arrives before the pressure one
At 150 bar a DN100 seat sees 118 kilonewtons of thrust, and that number is the same for every seat material. What changes with temperature is how much of it the material can take. PTFE at 185 degrees C is inside its continuous temperature range and well outside its comfortable compressive range: its allowable differential pressure at that temperature is a fraction of the ambient figure, and its creep means that a seat which sealed on the test stand will have taken a permanent set by the first maintenance interval. PEEK is better and still creeps under 15 megapascals at this temperature. Above 200 degrees C the argument is over and the seat is metal.
Metal seats do not seal bubble tight, which in this loop is acceptable for isolation and unacceptable for a leakage class written as a word. The number has to be stated at temperature and in both directions, because the seat that holds upstream pressure and the seat that holds downstream pressure are loaded by different geometry, and in a trunnion valve with a self relieving seat they may not both be loaded at all.
There is a trade that buyers should see written down. A piston balanced seat removes most of that 118 kilonewtons from the seat and makes the valve operable with a small actuator, and it does so by introducing a second cavity with its own stagnant liquid and its own oxygen problem. The unbalanced seat needs a bigger actuator and a heavier stem and has one less place for the fluid to sit. Neither is wrong. Choosing without knowing which one you chose is how a plant ends up with a valve that is easy to operate and impossible to keep.
8. What the licensor writes that the valve standard does not
ASME B16.34 will tell you the minimum wall for a Class 900 body at 185 degrees C and it will tell you nothing else that matters here. The document that governs this valve is the urea licensor's material and corrosion specification, and it is short, specific and unforgiving. It names the alloy by its own designation rather than by a UNS number, it caps ferrite on the finished component, it requires the weld procedure to be qualified on that alloy with that filler, it requires the solution anneal, and it requires a corrosion allowance of zero.
Zero allowance is the clause that changes how the valve should be bought. Every other body in the plant has a millimetre or three of sacrificial wall, so a pit is an Inspection finding with a budget attached. Here there is no budget: the alloy is not supposed to corrode at all, so any measured loss is a deviation from the design premise, and a four millimetre pit is not a wall loss to be monitored, it is a body replacement, because there is no allowance for it to consume. That is why the wall thickness map at the seat region, taken at construction and filed with the tag number, is worth more than any certificate in the package.
Two procurement consequences follow. First, positive material identification should cover the body, the ball and the trim, at 100 percent, because the failure mode is a compliant certificate attached to a non compliant part. Second, the ferrite reading belongs in the inspection and test plan as a hold point with a number, taken on the finished machined surface, not as a note in the material certificate. A supplier who cannot produce that reading has not made a urea grade valve; it has made a stainless valve and described it as one.
9. Seven clauses and one calculation for the purchase order
The clauses are ordinary once the mechanism is understood. Name the duty by the fluid and the seat temperature, not by the line class. Specify urea grade to the licensor's designation, with a finished part ferrite limit stated as a number, 100 percent positive material identification and a qualified weld procedure with the solution anneal witnessed. State the cavity volume, the relief direction and whether the cavity can be drained, vented and flushed. Specify the seat as metal with a leakage class at temperature in both directions, with the test medium named. Acknowledge the zero corrosion allowance in writing and require the wall thickness map. Require a warm demineralised wash connection of 25 millimetres on both sides of the seat, positioned so that the valve drains. And require the drawing to show where the liquid sits when the valve is closed, because that is the volume that will decide the inspection result in year two.
The calculation is the one already done above, and it belongs in the data sheet as a note rather than in a textbook: cavity volume, times density, times dissolved oxygen, divided by the daily oxygen demand of the wetted area at the design corrosion rate. It comes out in days. When the answer is two and a half days against a turnover of a hundred and forty, the designer has a reason to make the cavity smaller, and the buyer has a question to ask that no supplier can answer with a catalogue.
10. Where J-VALVES fits
J-VALVES ships into both ends of this loop: the forged high pressure isolation valves in urea grade stainless with a measured ferrite figure on the certificate, and the melt line valves with a jacket or a trace, a stated setpoint and a body geometry that drains. Bores run from DN15 to DN1400 and classes from ASME 150 to 2500, which covers the whole pressure staircase of a urea plant from the letdown station to the prilling feed. A body is proved with water at 150 percent of its rating on the shell and 110 percent on the seat, and each certificate records which side the test pressure was applied from, because in a loop with two different seat loadings that is not a detail. Every drawing shows the cavity, the drain and the vent, and every quotation names the alloy by the licensor's designation rather than by a trade name. Send the loop pressure, the seat temperature and the licensor's material clause, and a priced reply with the ferrite limit and the cavity volume written into it comes back the same working day, usually inside three hours.
11. Frequently Asked Questions
Why does a urea plant inject air into the carbon dioxide feed?
Because the corrosion resistance of the stainless steel in the high pressure loop is not a property of the alloy, it is a maintained electrochemical state. Carbamate removes the chromium oxide film faster than the metal can rebuild it unless oxygen is present, so 0.25 to 0.6 percent air by volume is added to hold every wetted surface in the passive region. Without it the corrosion potential drops into the active range and metal loss rises from 0.01 to 0.05 millimetres a year to 0.1 to 1.0, which is a factor of ten to fifty. The air is a corrosion inhibitor and should be treated as one.
Why did the valve bore stay bright while the cavity behind the seat pitted through?
Because the bore is swept and the cavity is not. A DN80 cavity holds about 0.4 litres, and the only exchange with the flowing stream is leakage past the seat, so at a few millilitres a day the liquid in it is roughly four months old. The dissolved oxygen in that liquid, around 44 milligrams, is consumed by the wetted area in two or three days at an active corrosion rate of 0.1 millimetres a year. The film therefore dies in the first week and the cavity corrodes actively for the rest of the year, in a stagnant pocket that acidifies itself, which is why the loss is a pit rather than general thinning.
Is 316L acceptable for ammonium carbamate service?
Only in its urea grade form, which is a different specification wearing the same name. Standard 316L carries three to eight percent delta ferrite, and carbamate attacks ferrite selectively along the stringers because it is a chromium and molybdenum depleted phase in galvanic contact with austenite. Urea grade raises nickel to about 14 percent, holds carbon at 0.02 or below and caps ferrite at 0.6 percent or lower, with fully austenitic weld metal and a solution anneal at 1,040 to 1,080 degrees C. Duplex 2205 and titanium are both offered as upgrades and both are wrong here.
How much gas does one litre of carbamate release when the pressure drops?
About two hundred litres at 5 bar. A litre of melt at 1,150 kilograms per cubic metre is 1.15 kilograms, or 14.7 moles of ammonium carbamate, and each mole dissociates into three moles of gas, so 44 moles, which is 986 litres at atmospheric pressure and 197 litres at 5 bar absolute. That expansion is why the letdown position needs a dedicated device with an expanded outlet and hardened trim, and why an isolation ball valve should never be used to throttle it.
What is the difference between the high pressure loop valve and the melt line valve?
The loop valve fails because it is too cold and too stagnant, and the melt valve fails because it is too hot. In the loop, 185 degrees C and 150 bar with a cavity that loses its oxygen is the danger. After the evaporators the fluid is a 99.7 percent melt at 140 degrees C with a freezing point of 132.7, so seven kelvin is the whole margin, and anything that holds the wall above 150 degrees C starts making biuret, which is the impurity that decides the fertilizer grade. Steam at 4 bar is 151 degrees C and steam at 6 bar is 165; the difference between those two jacket pressures is a product quality decision, not a utility one.
