A desalination plant on the Gulf coast replaced a 16 inch high pressure header valve fourteen months after commissioning. The specification called for super duplex, the certificate said 2507, the mill test report was in order, and the valve had passed its hydrotest at the works. When the body was cut out, there was a pit through the wall behind the downstream flange gasket, about the size of a fingertip, surrounded by a ring of rust coloured deposit. The failure was not a material substitution. It was a crevice, warm chlorinated seawater, and a gasket that had been wet for a year.
Seawater reverse osmosis is the most chemically aggressive continuous duty that a general purpose ball valve ever sees, and it is usually specified by people who are very good at membranes and pumps and less interested in valves. That is understandable, because the valve is a small part of the capital cost, and it is a large part of the availability problem. This article is about the high pressure section specifically: what the fluid actually is at 65 bar, why the material that works on the low pressure side is not automatically right here, and which of the failures are arithmetic that can be designed out.
1. The high pressure section is a chemical plant running at 65 bar
It helps to stop thinking of the high pressure train as plumbing. Between the high pressure pump and the membrane inlet, the fluid is seawater that has been dosed, filtered to a few microns, dechlorinated, and pressurised to a value that is set by the osmotic pressure of the concentrate rather than by any mechanical convenience. Everything downstream of the membrane is brine at roughly twice the salinity of the sea, still at pressure, and it has to be brought back to atmospheric without destroying the valve that does it.
The three distinct duties are worth separating because they need different valves. First, the suction and discharge of the high pressure pump, where the concern is pressure pulsation, NPSH and a valve that is usually fully open. Second, the membrane feed header and the per-rack isolation valves, where the concern is tight shutoff at 65 bar with a differential that reverses when a rack is taken out of service. Third, the brine energy recovery loop and the brine letdown, where the concern is a fluid at 65 bar that must lose that pressure in a controlled way, and where the valve is the place the energy is dissipated.
2. Recovery arithmetic: 35 grams per litre becomes 64
The number that drives everything is the recovery ratio. A plant recovering 45 percent of the feed sends 55 percent of it to drain, so the salt that came in with the feed leaves in a smaller volume, and the concentration factor is one divided by one minus the recovery, which is 1.82. Feed at 35,000 milligrams per litre therefore leaves as brine at about 63,700. Run the recovery to 50 percent and the factor is 2.0, and the brine is at 70,000. That is not a small change in a materials selection: it is a step change in chloride activity, in conductivity, and in the osmotic pressure the pump has to beat.
Osmotic pressure scales roughly with concentration, at about 0.7 bar per 1,000 milligrams per litre at ambient temperature for seawater. Feed at 35 grams per litre therefore has an osmotic pressure near 26 bar, and the concentrate at 64 grams per litre is near 47 bar. The pump has to supply the concentrate osmotic pressure plus the net driving pressure needed for flux plus the pressure losses, which is why a modern SWRO train runs at 60 to 70 bar and a high recovery or high salinity plant runs above that. It is also why a valve rated at 40 bar because somebody read the pump curve at the wrong point is a valve that will be replaced.
Temperature matters more than most specifications admit. Membrane permeability falls by roughly 3 percent per degree Celsius, so a plant designed at 25 degrees C needs about 3 percent more pressure for every degree colder, and a winter feed at 15 degrees C can push the same train from 65 to 78 bar. A valve selected on the summer number is a valve operating outside its design envelope for four months a year, and the seat load and the operating torque go with it.
Parameter | Typical value | What it does to the valve |
|---|---|---|
Feed total dissolved solids | 35,000 mg/L, open intake | Sets the chloride load on every wetted surface |
Recovery | 45 percent, single pass | Concentration factor 1.82, brine at about 64,000 mg/L |
Feed osmotic pressure | about 26 bar at 25 degrees C | Lower bound of the pressure the valve must hold |
Concentrate osmotic pressure | about 47 bar | Explains why the pump, and the valve, see 65 bar and not 30 |
High pressure header | 60 to 70 bar, up to 80 in cold feed | Class 600 is the normal starting point, not Class 300 |
Brine discharge to outfall | 1 to 2 bar | 63 bar must be dissipated, and a single valve cannot do it |
Free chlorine upstream of membranes | below 0.1 mg/L after dechlorination | Residual oxidant is the difference between passive and pitting |
Specific energy | 3.0 to 4.0 kWh per cubic metre with energy recovery | Every bar of valve pressure loss is paid for continuously |
3. Why 316 is wrong here, and why 2507 is not automatically right
Type 316 stainless in seawater is a scheduled failure. Its pitting resistance equivalent number, calculated as chromium plus 3.3 times molybdenum plus 16 times nitrogen, is around 25, and the accepted threshold for reliable seawater service is above 40. Between 25 and 40 there is a zone occupied by the super austenitic grades such as 254 SMO and the 6 percent molybdenum alloys, which work in some seawater duties and not in others, and above 40 sit the super duplex grades such as 2507, with a PREN near 43, and the hyper duplex grades above 48.
2507 is the workhorse of the SWRO high pressure section for good reasons: a yield strength around 550 MPa, which lets a Class 600 body be compact, good resistance to chloride stress corrosion cracking, and a thermal expansion close enough to carbon steel that a transition joint is straightforward. What it does not have is immunity. Duplex and super duplex are vulnerable to crevice corrosion and to hydrogen embrittlement, and both of those are failure modes that a pressure test at the works will never reveal.
The crevice issue is the one that kills valves, because a valve is a machine made of crevices. Every gasket face, every seat pocket, every O-ring groove, every threaded plug and every flange joint is a gap where the oxygen is consumed faster than it can be replenished, and where the local chemistry drifts to low pH and high chloride. A super duplex body that is perfectly passive in the bore can be actively corroding under the bonnet gasket, and the pit grows inward, invisible, until it is through the wall.
4. The crevice temperature ceiling, and the chlorination trap
Super duplex has a critical crevice temperature in natural seawater that is generally quoted between 40 and 50 degrees C for 2507, and lower for the 6 percent molybdenum austenitics. Below that temperature the alloy repassivates in a crevice; above it, it does not. This matters because a SWRO high pressure valve does not sit at seawater temperature. The pump adds 3 to 5 kelvin, the feed can be warmed, and a valve in a plant room in a hot climate sits in still air at 40 degrees C with a surface that is warmer than the fluid.
The second variable is oxidant. Chlorination of the intake, typically 0.5 to 2 milligrams per litre of free chlorine with dechlorination by sodium metabisulphite before the membranes, is standard practice for biofouling control. But the dechlorination point is not always upstream of every valve, and during a dosing upset, a shock dose, or a cleaning cycle, the high pressure section can see oxidant. Free chlorine raises the corrosion potential of stainless steel by several hundred millivolts, from around minus 200 millivolts against a saturated calomel electrode in quiet seawater to plus 200 or more, and that shift is exactly what pushes a duplex alloy from the passive region into crevice attack at a temperature where it would otherwise have been safe.
The practical consequences are three. Do not specify a valve for the high pressure section without knowing where the dechlorination point is relative to it. Do not accept a design that has unnecessary crevices, such as a two piece body with a gasket in the wetted path where a welded or monobloc body would do. And specify a corrosion test rather than a composition: ASTM G48 Method A in ferric chloride at an agreed temperature, typically 40 degrees C for 2507, with a maximum weight loss and no visible pitting, is a test of the actual heat treatment and surface condition, which is what determines whether the crevice resistance is there.
5. Velocity has two limits, and they point in opposite directions
Seawater valves are usually selected on a velocity rule of thumb, and the rule of thumb is quoted without its two ends. The lower limit exists because slow seawater grows things. Below about 1 metre per second, and certainly in a valve that sits closed or partly open for long periods, mollusc larvae and barnacles settle, sulphate reducing bacteria establish in the deposit, and the crevice under the deposit becomes the corrosion site. A dead leg in a seawater line is a corrosion coupon with a biofilm on it.
The upper limit exists because fast seawater erodes the protective film and, with the suspended sand that a beach well or an open intake delivers, mechanically removes the metal. For super duplex in brine the erosion-corrosion threshold is generally taken as 5 metres per second for clean service and somewhat lower with sand, and the practical design band is 1.5 to 3.5 metres per second in the headers with a full port ball valve, which is one of the reasons a full port valve is specified rather than a reduced port one: the bore velocity, not the pipe velocity, is what the seat and the ball actually see.
In a ball valve specifically, the geometry matters more than the nominal velocity. When the valve is throttled, the flow accelerates through the gap between the ball and the seat and can reach several times the pipe velocity at that gap, with a jet that impinges on the downstream body wall. A ball valve is not a control valve, and using one at 30 percent open in continuous service in seawater is a way to destroy both the seat and the body in a year.
6. Energy recovery puts a pressure ripple through the whole train
Modern SWRO plants recover the pressure energy in the brine with a pressure exchanger, an isobaric device in which the high pressure brine pressurises a slug of low pressure feed directly, with an efficiency of 95 to 97 percent and a lubrication flow of 1 to 3 percent of the throughput. The device is a rotating or reciprocating rotor with ceramic or duplex ports, and its output is not perfectly steady: the pressure at the high pressure outlet carries a ripple at the rotor frequency, typically a few hertz, with an amplitude that depends on the number of chambers and the control scheme.
The ripple finds every check valve, every flow meter and every isolation valve on the header. Its consequences are fatigue in the seat and the packing, chatter in a check valve, and noise. For the ball valve, the relevant question is whether the valve is expected to modulate. It is not. A ball valve in this position should be fully open or fully closed, with the actuator sized for the differential at the moment of opening, which is 65 bar across a 400 millimetre ball, not the running torque.
That is also the argument against using the isolation valve for flow control during a rack flush or a membrane clean. Every one of those operations is done better with a dedicated control valve, a needle valve or an orifice plate, and the ball valve kept as a barrier. A ball valve that is asked to throttle in brine will cavitate, wear and leak, and the leak will be into a membrane that cannot tolerate a pressure spike.
7. Scale grows in the slow places, which is inside the valve
Concentrate is supersaturated with respect to calcium carbonate, and often with barium sulphate and calcium sulphate, which is why antiscalant is dosed at 2 to 5 milligrams per litre and why the concentrate is kept above a minimum velocity. Scale does not precipitate uniformly in a pipe. It deposits where the supersaturation is highest, the temperature is highest, and the flow is slowest, and a valve cavity is all three: a dead volume that exchanges slowly with the main flow, a body that is warmer than the fluid, and a seat pocket where the local velocity approaches zero.
The result is a valve that still turns but no longer seals, because the seat face has a hard rim of calcium carbonate on it, and the ball has a matching band. The damage is done when somebody closes the valve on that scale and the elastomer is cut. On a plant with a scaling excursion, the isolation valves are often the first casualty and the last to be investigated, because the membranes show the symptom.
The countermeasures are unglamorous. Keep the cavity flushed by leaving the ball in the open position in normal service, which also protects the seat from being loaded against a static ball for months. Provide a drain or a flush port on the cavity where the design allows, and use it on the maintenance schedule rather than only at commissioning. Do not accept a valve with a large unswept cavity in the body, because the cavity is where the first scale forms and it is also where the first corrosion cell starts.
8. Brine letdown: 63 bar through one valve is a cavitation machine
Brine leaves the membrane at essentially the feed pressure minus the losses. It has to arrive at the outfall at 1 or 2 bar. That is 60 bar or more of pressure energy, and it has to be removed somewhere. If it is removed in a single valve, the valve is a cavitation device by construction: the fluid accelerates through the throttling gap, the pressure falls below its vapour pressure, vapour bubbles form, and they collapse downstream against the body wall with a local pressure spike that removes metal and elastomer.
The standard answer is to break the letdown into stages, typically three, with roughly 20 bar per stage, and to use a valve designed for the duty: an angle body, a sleeve or a multi-stage trim, with the flow directed at a hardened target rather than at the body wall. Where a ball valve must be used for isolation in that line, it is a barrier and not a throttling element, and it should be a metal seated valve with a hardened ball and seat if it will ever see cavitating flow.
The numbers help. Cavitation inception is governed by the cavitation index, sigma, equal to the downstream pressure minus the vapour pressure divided by the pressure drop. For water at 25 degrees C the vapour pressure is 0.03 bar, so a valve dropping 65 bar to 2 bar has a sigma near 0.03, which is far below the 0.2 to 0.4 at which most valves begin to cavitate, and in the region where choked cavitation is fully established. No seat material survives that in continuous service. Choked cavitation also caps the flow: beyond the choked point, opening the valve further does not increase the flow, it only increases the damage.
9. Surge: the 17 bar that arrives when the valve closes
A high pressure brine line is a stored energy device, and a fast closing valve converts that energy into a pressure wave. The magnitude is given by the Joukowsky relation, the density times the wave speed times the change in velocity. For seawater at 1,025 kilograms per cubic metre, a wave speed of about 1,100 metres per second in a steel pipe, and a velocity change of 1.5 metres per second, the surge is 1.7 megapascals, or about 17 bar on top of the steady 65.
Whether the full surge develops depends on the closing time compared with the pipe period, twice the length divided by the wave speed. For a 100 metre line that period is about 0.18 seconds. A valve that closes in half a second is therefore closing slowly relative to the wave and the surge is attenuated; a solenoid trip that slams a ball valve shut in 0.1 seconds develops the full 17 bar, and a check valve that slams can do worse. On a Class 600 line at 65 bar, 17 bar of surge is a 26 percent overpressure, which is within the hydrotest but well outside the daily fatigue envelope of a gasket and a seat.
The design answers are a slow closing actuator with a flow control on the exhaust, a surge relief valve or a rupture disc on the header, and a flywheel or a soft starter on the pump so that the pump does not stop before the valve closes. All three are cheaper than the membrane that gets a pressure spike through a leaking seat.
Symptom in service | Mechanism | What it usually means | Action |
|---|---|---|---|
Pit through the body behind a gasket | Crevice corrosion, warm chlorinated seawater | Oxidant present, crevice geometry, temperature near the ceiling | Move the dechlorination point, remove the crevice, verify alloy by test |
Seat leaks only after a cleaning cycle | Scale rim broken off, or oxidant attack on the elastomer | Flush not performed, or chemical clean reached the seat | Inspect and dress the seat, replace soft goods, add a flush step |
Valve will not open against full differential | Actuator sized on running torque, not breakout torque | Torque calculation used the wrong case | Re-size the actuator, add a bypass, check the seat relief arrangement |
Erosion groove downstream of the seat | Throttling a ball valve in brine, or sand | The valve is being used as a control valve | Stop throttling, add a control valve, consider a hardened trim |
Noise and vibration on the brine line | Cavitation at letdown, or ERD ripple | Single stage letdown, or a resonance in the header | Stage the letdown, review support and the ERD control |
Gasket weep after a pump trip | Surge overpressure and subsequent relaxation | Fast closing valve, no surge protection | Slow the closure, add relief, re-torque to procedure |
10. Seats, seals and the membrane next door
The soft goods in a SWRO valve are a shorter list than in most services, and the choice is narrower than catalogue tables suggest. EPDM is the default for seawater: it tolerates hypochlorite and ozone far better than nitrile, it handles the temperature band, and it is compatible with the antiscalants and the cleaning chemicals. PTFE and its modified grades are used where the temperature or the chemical resistance demands it, with the usual caveat that PTFE creeps and permeates, and that a PTFE seat in a high differential application needs a proper seat support rather than a free ring.
Nitrile, which is the cheapest and most common seat material in general service, is the wrong choice here for a reason that is often missed: it degrades under residual chlorine and under ozone, and the degradation product is a sticky film that then attracts scale. A nitrile seat in a chlorinated seawater line can fail in a way that looks like a scaling problem.
The membrane is the reason the cleanliness requirements are stricter than for any other water service. A reverse osmosis membrane is a thin polyamide layer that is damaged by free chlorine, by oil, by grease and by fine particulate, and it is a few metres downstream of the valve. A valve that is packed with a hydrocarbon grease, hydrotested with unfiltered water, or assembled with a silicone lubricant on the seat will deliver that contamination to the membrane. The specification for the high pressure section should therefore say: no hydrocarbon lubricant on any wetted part, water for hydrotest filtered to 5 microns and free of chlorine, and the bore cleaned and passivated after any welding or machining.
11. Testing, cleanliness and the paper that decides a claim
A hydrotest proves the body does not leak. It proves nothing about crevice resistance, nothing about whether the correct heat treatment was applied, and nothing about whether the valve will still be passive in a year. For SWRO the test package should carry three things beyond the standard shell and seat test: a corrosion test on the body material to ASTM G48 Method A at an agreed temperature, a positive material identification check on the body, the ball and the trim, and a ferrite check on the duplex, because a duplex weld or heat treatment that produces the wrong ferrite balance loses both toughness and corrosion resistance while keeping the same certificate.
The ferrite number is worth a paragraph because it is invisible. Super duplex is a two phase alloy, roughly half austenite and half ferrite, and the balance is set by composition and by cooling rate through the solution anneal. Too much ferrite and the alloy becomes susceptible to hydrogen embrittlement and to selective attack; too little and it loses strength. The acceptable range for 2507 is typically 35 to 55 percent ferrite, measured on the finished part, and a body that was annealed incorrectly can sit outside that range with a mill certificate that looks perfect because the certificate describes the heat, not the part.
Finally, the paper. A claim about a seawater valve is settled by the heat number traceability, the corrosion test report, the ferrite measurement and the hydrotest record, in that order, and a supplier who cannot produce all four is a supplier who will argue. The J-VALVES works covers a size envelope from DN15 up to DN1400, which in imperial terms is 1/2 inch through 56 inch, and a pressure envelope from ASME Class 150 to Class 2500, with body materials including carbon steel, stainless, duplex, super duplex and nickel alloys across a temperature band of minus 196 to plus 650 degrees C. Shell proof is 1.5 times the cold working pressure and seat proof 1.1 times, to API 598 or API 6D as the purchase order requires, with G48 and ferrite records available for seawater scope, and quotations are issued inside three working hours.
12. A specification clause for the high pressure section
Valves in the high pressure section shall be full port, trunnion mounted, with a body in super duplex stainless steel UNS S32750 to ASTM A182 or A995 as applicable, solution annealed and water quenched, with a ferrite content between 35 and 55 percent verified on the finished body, and shall be tested to ASTM G48 Method A at 40 degrees C with no visible pitting and a weight loss not exceeding 4 grams per square metre. Body construction shall minimise crevices in the wetted path, and where a gasketed joint is unavoidable the gasket shall be confined and the crevice shall be excluded from the flow path by design. All wetted surfaces shall be free of hydrocarbon lubricant, and hydrotest water shall be filtered to 5 microns, free of chlorine, and drained and dried after test. Soft seats shall be EPDM or a PTFE based material as specified, and nitrile shall not be used. Actuators shall be sized on the breakout torque at the maximum differential including a 1.5 factor, with a closing time not shorter than the pipe period unless surge protection is provided. Each valve shall be supplied with a 3.1 certificate, a positive material identification record, a G48 report, a ferrite measurement record and a cleanliness certificate, and shall be packed with the bore dried, capped and preserved.
None of that is unusual in a well run desalination project, and all of it is routinely missing from a valve package bought on price. The cost of the valve is small compared with the cost of a membrane rack, and the difference between a fourteeen month life and a fifteen year life is usually a gasket detail, a test report and a dechlorination point on a P and ID that nobody checked.
13. Frequently Asked Questions
What material should a SWRO high pressure ball valve be made from?
Super duplex stainless steel UNS S32750, with a pitting resistance equivalent number around 43, is the normal choice for the high pressure header and the brine loop. Type 316 with a PREN near 25 fails in seawater, and the 6 percent molybdenum austenitics sit in between with a lower crevice temperature. The material callout is only half the answer: the body must be solution annealed and water quenched, the ferrite content must be verified at 35 to 55 percent on the finished part, and the crevice resistance should be proven by an ASTM G48 Method A test rather than inferred from a mill certificate.
Why do super duplex valves pit under gaskets and not in the bore?
Because a crevice is a geometry, not a material. Under a gasket the oxygen is consumed faster than it can be replenished, the local chemistry drifts to low pH and high chloride, and the alloy can no longer repassivate. Super duplex has a critical crevice temperature in natural seawater of roughly 40 to 50 degrees C, and a plant room in a hot climate plus a pump that adds a few kelvin can put the metal at that limit. Free chlorine from an intake chlorination system raises the corrosion potential by several hundred millivolts and makes the attack possible at temperatures that would otherwise be safe.
How much pressure does a SWRO high pressure pump actually need?
Enough to overcome the osmotic pressure of the concentrate plus the net driving pressure for flux plus the losses. Feed at 35,000 mg/L has an osmotic pressure near 26 bar, and at 45 percent recovery the concentrate is at about 64,000 mg/L with an osmotic pressure near 47 bar, which is why the header runs at 60 to 70 bar. Temperature adds to it: membrane permeability falls by about 3 percent per degree Celsius, so a train designed at 25 degrees C may need 78 bar at 15 degrees C. Valves and flanges should be selected on the cold case, not the design case.
Can one valve handle the brine letdown from 65 bar to atmosphere?
No. A single valve dropping 65 bar to 2 bar operates at a cavitation index near 0.03, which is deep in choked cavitation, and the collapse of the vapour bubbles removes seat and body material within months. The standard arrangement is three letdown stages of about 20 bar each, using angle body or multi-stage trim valves with a hardened target, and the ball valve in that line is an isolation barrier rather than a throttling element. Choked cavitation also limits the flow, so opening the valve further adds damage without adding capacity.
How fast can a high pressure brine valve close before surge becomes a problem?
Compare the closing time with the pipe period, twice the length divided by the wave speed. For a 100 metre line with a wave speed near 1,100 metres per second the period is about 0.18 seconds, and a closure slower than that attenuates the surge considerably. A fast solenoid trip that shuts a ball valve in 0.1 seconds develops the full Joukowsky head, which for a 1.5 metre per second velocity change in seawater is about 17 bar on top of the steady pressure. Slow closure through an exhaust flow control, plus surge relief on the header, is the usual protection.
