J-VALVES · Ball Valve Manufacturer · Wenzhou, China info@j-valves.com +86-13600648865

API 6D ball valve selection guide

News

Home News Molten Salt Ball Valves: The Valve Works at 565 C And Dies at 219 C
Inquire
Metal seated ball valve with heat tracing and insulated flanges on a 565 C molten salt thermal storage header

There is a temperature at which a molten salt valve works and a temperature at which it fails, and on the instrument they sit 345 kelvin apart. The salt is liquid and useful between 220 degrees C, where the sodium and potassium nitrate eutectic freezes, and about 565 degrees C, where it begins to decompose into nitrite and oxides. Below the first number the valve is holding a solid. Above the second the fluid is attacking its own container. Everything a molten salt valve does happens inside that gap, and the gap is not generous.

What separates this from ordinary high temperature duty is that the valve is never uniformly hot. A valve in 500 degree steam is hot on both faces of every wall and the problem is strength. A valve in a salt header is hot in the bore and cold at the handwheel, and the problem is the boundary between the two: the flange that is 40 degrees outside and 560 inside, the stem that carries heat out of the body, the cavity that traps a liquid which grows by nearly a tenth of its volume across the working range. Nearly every failure in this service happens at a boundary rather than in a body, and it is usually written up in the maintenance system as seat damage, which is the symptom and not the cause.

The rest of this article is the arithmetic behind that claim. It covers what the salt actually is in numbers, why the same stored energy needs almost three times the tonnage at 386 C that it needs at 565 C, why a heat-up is a ratcheting question rather than a strength question, why a liquid-full cavity here asks the body for more than a thousand bar, why freezing contracts instead of bursting, and why the seat leakage that matters is the leak that freezes somebody else's branch eight hours later.

1. Two temperatures, and a gap that sets every material choice

The eutectic used in almost every commercial plant is 60 percent sodium nitrate and 40 percent potassium nitrate by weight. It melts at 220 C and is stable to roughly 565 C, above which the nitrate ion reduces to nitrite, nitrogen oxides come off, and the fluid becomes both corrosive and chemically different from the one that was specified in the purchase order. Neither number is a valve property, and together they are the only two numbers the valve design truly has to respect.

Between them the fluid is unremarkable hydraulically. Viscosity runs about 3 centipoise at 300 C and under 2 at 500, within a factor of three of water, so a flow coefficient carried over from water service gets the sizing close enough. Specific heat is 1.5 kilojoules per kilogram kelvin, conductivity is around 0.5 watts per metre kelvin, and it wets steel well enough that a dry seat is not a realistic starting condition after the first month. The difficulty is not the flow. The difficulty is that all of those numbers have to hold at a temperature where the steel has lost half its room temperature strength and every sealing polymer in the catalogue has already been eliminated.

Property

Value across the working band

What it decides on the valve

Freeze point of the 60/40 eutectic

220 C

Trace heating setpoint, cavity geometry, seat clearance

Decomposition onset

about 565 C, rapid above 600 C

Upper limit for every wetted alloy, packing and gasket

Liquid density

1,950 kg/m3 at 220 C, 1,730 at 565 C

Flow coefficient, cavity pressure, tank volume

Volumetric expansion

3.5e-4 per K, nearly 9 percent over 275 K

Relief direction, vent size, spring rate

Specific heat

1.5 kJ/kg K

Storage tonnage, standing heat loss, trace heating bill

Viscosity

3 cP at 300 C, under 2 at 500 C

Behaves like water in the bore, not like thermal oil

Heat of fusion

about 160 kJ/kg

What a frozen plug costs to melt, and how long it takes

Solid density at the freeze point

about 2,100 kg/m3

Freezing contracts, so nothing bursts and nothing seals

2. Nine percent of a trapped volume has nowhere to go

Density falls from about 1,950 kilograms per cubic metre at the freeze point to about 1,730 at 565 C. That is 3.5 times ten to the minus four per kelvin, and 9 percent across the 275 kelvin working band: an unremarkable number for a liquid, and the single most important number in the valve.

Multiply the expansion by the bulk modulus, around 1.6 gigapascals for a nitrate melt, and a rigid liquid-full cavity gains 5.6 bar for every kelvin. The working band is 275 kelvin. The product is 1,540 bar, which no Class 300 body will hold, so the fact that these bodies do not burst in service is proof that the cavity is never perfectly rigid and that something lifts. What lifts is the seat, and whether it lifts in the intended direction is settled on the drawing rather than in the field.

This is where a specification copied from ordinary liquid service goes wrong. A self-relieving seat is normally written as a feature that protects the body from a trapped liquid expanding as it warms, and the only question asked is whether it relieves. In molten salt the cavity also has to be able to draw liquid back in, because the common case is the reverse one: the valve is closed hot, the line cools, the trapped salt contracts, and the cavity is in vacuum within a few kelvin. A cavity that has cavitated is no longer liquid full, and the next heat-up compresses the void and then locks the cavity hydraulically against a seat that the vacuum has already pulled shut.

The arrangement that works is a cavity vent on the downstream side sized for salt and not for air. Twenty five millimetres is the useful minimum, because anything smaller is plugged by a frozen film during one shutdown, and the vent needs its own trace heating at the same setpoint as the body. A ten millimetre vent with a thermostat is a vent that works until the first cold night of the year.

3. The same gigawatt hour, at almost three times the tonnage

One thousand megawatt hours is 3.6 million megajoules. Divide by the specific heat and the working band, 1.5 kilojoules per kilogram kelvin times 275 kelvin, and a tower plant holding salt between 290 and 565 C needs 8,700 tonnes of it. The same calculation on a parabolic trough plant, where the heat transfer oil limits the salt to 292 to 386 C and the band is 94 kelvin, gives 25,500 tonnes. Andasol 1 carries 28,500 tonnes for its 1,010 megawatt hours, which is the same number plus the heel that never leaves the tanks.

That factor of three is the whole reason the tower configuration won the last decade of concentrated solar. It is also the reason a valve maker sees two different customers. The trough plant buys a great many valves at a moderate temperature and cares about price, spare parts and the cost of replacing a seat in a tank farm. The tower plant buys fewer valves at a temperature where the seat, the packing and the trace heating circuit are all custom, and cares about whether the thing survives ten thousand cycles.

The mass also fixes where the valves are. Salt leaves the hot tank at 565 and returns to the cold tank at 290, so the two banks of isolation valves on either side of the steam generator see the same fluid at temperatures 275 kelvin apart, with the same body drawing and different seat, gasket and trace heating specifications. Buying them as one line item is the most common procurement mistake on these plants, and it shows up three years later as a cold side bank that cannot be isolated because every seat in it was qualified at the hot end.

4. A heat-up is a ratcheting question, not a strength question

Thermal stress from a through-wall gradient is roughly the elastic modulus times the expansion coefficient times the difference, divided by two. For austenitic stainless that collapses to 1.6 megapascals per kelvin, which is a number worth memorising because it converts any temperature difference into a stress without a spreadsheet. At 565 C the yield strength of 316 is around 130 megapascals, so the elastic budget is about 80 kelvin of through-wall difference. Under 80 K the wall bends and springs back. Over 80 K, every cycle leaves plastic strain behind and the life of the part is being spent.

A cold start is nowhere near 80. The body sits at 290 on trace heating, 565 salt arrives, and the inner surface sees a 275 kelvin step against a wall 13 millimetres thick. The surface stress for that step is well over a gigapascal, several times yield, applied within the first second. The structure does not fail. It yields a thin surface layer, locks in a compressive residual stress, and thereafter behaves elastically for cycles of the same magnitude. That is why the first heat-up of a new valve matters more than the thousandth, and why a slow first heat-up is not a courtesy to the equipment but the event that sets the residual field for the next thirty years.

The useful conclusion is not "heat up slowly", which everyone already believes. It is that the trace heating setpoint is a thermal shock parameter and not only a freeze protection parameter. A body held at 250 C before the first salt arrives is 315 kelvin from the fluid and four times over the elastic budget. A body held at 500 C is 65 kelvin away and inside it, with the same heater and a different thermostat setting. The cost of the higher setpoint is a few hundred watts of standing loss; the cost of the lower one is a plastic cycle on every start, and nobody counts those against the creep-fatigue allowance because the valve does not fail for eight years.

5. The wall is four times thicker than the pressure needs

Do the thin-wall arithmetic once, because it changes where the engineering effort belongs. At 20 bar, a 219 millimetre bore and an allowable stress of 75 megapascals at 565 C, the required wall is 2.9 millimetres. The flange standard asks for about 13. Hoop stress in the 13 millimetre wall is 17 megapascals, a fifth of the allowable and a fifth of the hundred thousand hour rupture strength at that temperature, so neither pressure nor creep is sizing this body. What sizes it is the class rating that comes with the flange pattern, casting and forging practice, and the wish to have something stiff enough to hold a seat flat after ten thousand cycles.

Two things follow. The first is that nobody wins this application by adding wall. A molten salt valve is won in the seat, the stem, the packing and the trace heating circuit, and the body is a commodity whose real jobs are to be stiff and to have enough thermal mass that it does not follow the salt down when the loop stops. The second is that the pressure class is a poor guide to anything. A Class 300 salt valve is a 20 bar valve inside a 51 bar casting, and quoting it as a 51 bar valve because the tag says 300 is the kind of error that survives until the first hydraulic analysis.

6. Freezing contracts, and that is worse than bursting

Water is unusual in expanding when it freezes, which is why a frozen water line splits. Nitrate salt is normal. The solid at the freeze point is about 2,100 kilograms per cubic metre against 1,950 for the liquid, so freezing contracts the volume by roughly 7 percent. A frozen plug pulls away from the pipe wall and leaves an annulus, and a valve full of solid salt is not a valve under stress. Nothing bursts.

Three consequences follow and all of them are quieter than a burst. The first is that a frozen plug is not a plug: it does not seal, it does not hold pressure, and a freeze isolation that was going to buy a maintenance window passes salt through the annulus as soon as the upstream side warms. The second is that the frozen mass grips the ball. Solid nitrate shears at a few megapascals, and a ten millimetre annulus around a 250 millimetre ball is worth around 60 kilonewton metres of breakaway torque against a normal operating torque of one or two. That is thirty times the actuator, so the actuator stalls, the stem winds up, and the ball stays where it was with a twisted key and an operator who reports that the valve is stuck rather than frozen.

The third is that melting is not uniform. Heat arrives through the body, so the salt melts from the outside in and the last solid is a lump sitting on the seat, which is then crushed into the sealing surface on the first stroke. That is the mechanism behind most of the seat damage in this service, and it produces a scratch pattern that looks exactly like particle damage, which is why the investigation usually ends in the wrong place.

The operating rule that follows is short and needs to be written into the procedure rather than assumed: no stroke until the body is above 250 C along its whole length, measured at the seat rather than at a thermostat two metres up the pipe, and no valve is ever forced. A stalled actuator is information. A bar on the handwheel is the destruction of the seat.

7. Here the seat leak is a freeze, not an emission

Metal seats are mandatory above about 300 C, and metal seats do not seal bubble tight. Suppose the seat were parted by a uniform tenth of a millimetre, which is roughly what a ground seat looks like after a few thousand cycles at temperature. On a 200 millimetre seat circumference the leak area is 63 square millimetres. At 5 bar differential, a density of 1,800 and a discharge coefficient of 0.62, that passes 3 cubic metres per hour. Three cubic metres per hour is not a leak, it is a stream.

In most services that number is an argument about leakage classes. Here it is an argument about what happens downstream, and the answer catches operators out. The leaking salt does not freeze while it is leaking, because it arrives at 565 C and heats the pipe it runs into. It freezes an hour after somebody closes the upstream valve and walks away, in the coldest place, which is the flange at the end of the untraced branch. That branch then holds a solid plug which has contracted away from the wall, and the next time it is needed the flow goes around the plug and not through the line.

The specification that follows is unusual in three ways. The leakage class has to be a number rather than a word. The downstream line has to be traced whether or not the valve is expected to leak. And the seat test has to be run at temperature and in both directions, because a Rate D certificate issued at ambient on water says nothing about a ground Stellite seat at 565 C with a gradient through the body and a salt film on the contact band.

8. The stem is the coldest point, and the top of the ball comes next

A 200 millimetre stem of 316 running from a 565 C body to an actuator at 40 C grows by 1.8 millimetres. Packing compression absorbs 0.3 to 0.5, so the gland needs a live load with at least 3 millimetres of travel, and it needs it because the same 1.8 millimetres returns on every cool-down and packing does not recover from being over-compressed ten thousand times.

The stem is also a fin, and it is the reason the top of the ball is the first place salt freezes. A 30 millimetre stem 250 millimetres long draws about 24 watts out of the body into the yoke. That is nothing in energy terms and decisive in thermal terms: it holds the upper trunnion a few tens of kelvin below the bulk, and a few tens of kelvin is the entire margin on a body held at 250 C on a cold night.

Two details follow from that geometry. Graphite packing oxidises in air above about 450 C, and the top rings sit at the gland where the stem is hot and the air is not excluded, which is part of why the stem on a salt valve is long and the yoke is open rather than enclosed. And the cavity connections belong where the phase is: the vent at the top, because that is where a gas pocket collects during a drain, and the drain at the bottom, because that is where the solids end up.

9. Seven cases the drawing has to answer

Thermal design is usually presented as one case at one temperature. In this service the valve has seven lives, and the table below is the one to work through before the drawing is released, because each row is decided by a different feature.

Case

State of the salt

Metal

What the valve must survive

Where it is decided

Circulation, valve open

liquid, 565 C

565 C throughout the wetted path

Creep relaxation of the seat spring, ball growth against the seat

Energiser alloy, cold clearance

Charge from the 290 C tank

liquid, 290 to 565 across the body

gradient through the wall

275 K step on the inner surface

Trace heating setpoint, ramp control

Closed valve, loop running

liquid trapped, both ends sealed

565 C

5.6 bar per kelvin in a cavity with no exit

Relief direction, vent bore

Shutdown, valve closed

liquid cooling, then vacuum

down to 250 C on trace heating

Cavitation, then a seat pulled shut by the void

Bidirectional relief

Trace heating failure overnight

solid

25 C

Ball frozen in a salt ring, 60 kN m breakaway

Actuator sizing, no-force rule

Drain and maintenance

air in the bore, salt film on the seat

250 C

Film crushed on the first stroke after restart

Seat geometry, cleaning procedure

Ten thousand cycles, thirty years

liquid

cycling 290 to 565

Creep-fatigue of the contact band

Contact stress, hardness, lapping

Read the rows as a sequence of decisions rather than a list of hazards. The trace heating setpoint appears in row two and the vent bore in row three, and both are fixed before the first casting is poured. By the time the plant is commissioning, the only remaining freedom is the operating procedure, and an operating procedure cannot rescue a cavity that was drawn with one relief direction.

10. What the purchase order has to say

Seven clauses are enough to make this buyable and auditable.

First, name the salt, not just the temperature. The eutectic composition, the chloride limit, the perchlorate limit and the moisture limit belong on the data sheet, because at 565 C a contaminated batch does more damage to the wetted path than the alloy choice does.

Second, state the trace heating as a performance requirement with a wattage, a setpoint and a sensor location. A valve delivered with a heater pocket in the wrong place will be held at 250 C at the seat and 400 at the flange, and the seat is where the freeze matters.

Third, state the cavity relief in both directions and name the vent bore. A single relief direction is a cavity that cavitates on every shutdown.

Fourth, state the seat leakage as a number, at temperature, in both directions, with the test medium named. Ambient water is not a substitute and should not be accepted as evidence.

Fifth, specify the stem length and the live load travel, and require the packing to be qualified at the gland temperature that the stem length actually produces rather than at the line temperature.

Sixth, require the drain and vent connections at 25 millimetres minimum, traced, with the same setpoint as the body. This is the clause that gets value-engineered out and the one that produces the first winter failure.

Seventh, ask for the cold-start case to be witnessed. A valve that has been taken from 250 C to 565 C in one step, held, cycled and then leak tested is a valve whose seat has already met the worst thing that will happen to it. A valve tested only at steady state has been tested in its easiest condition.

11. Where J-VALVES fits

J-VALVES builds the hardware for both ends of that decision. Bores run from DN15 to DN1400, classes from ASME 150 through 2500, and a temperature band that reaches from 196 below zero to 650 above it, bracketing the salt loop with margin at each end. A body does not leave without a hydrostatic proof at 1.5 times its rating, and a seat does not leave without 1.1 times, each recorded with the side the pressure arrived from, because in this service the correct side is not always the obvious one. Metal seated trims are lapped and mapped rather than merely ground, the drawings show the relief path in both directions, and the material justification travels with the price. Send the salt chemistry, the two loop temperatures and the trace heating philosophy, and a priced offer with the reasoning attached is on its way back within 3 working hours.

12. Frequently Asked Questions

At what temperature does the salt freeze inside a valve, and what does it do to the valve?

The 60/40 sodium and potassium nitrate eutectic freezes at 220 C. Below that the valve is holding a solid which has contracted about 7 percent by volume, so nothing bursts and nothing seals either: the plug pulls away from the wall and leaves an annulus. The damage comes on restart, when the ball is gripped by a salt ring worth around 60 kilonewton metres of breakaway torque against a normal operating torque of one or two, and when the last solid to melt is crushed into the seat on the first stroke.

Why can a molten salt valve not use a PTFE seat?

Because every polymer seat in the catalogue has failed long before 565 C. PTFE is finished around 200 to 260 C depending on grade, PEEK around 300, and above that the seat has to be metal, normally a Stellite overlay on a stainless carrier. Metal seats do not seal bubble tight, which is acceptable in most high temperature service and is a specific hazard here, because leakage into an untraced branch freezes there an hour after the valve is closed.

Does a frozen salt plug burst the valve body?

No. Nitrate salt contracts about 7 percent when it freezes, unlike water, so a frozen plug separates from the wall instead of loading it. The consequences are different and quieter: the plug does not isolate, the ball is mechanically locked, and the melt is uneven so a lump of solid salt ends up on the sealing surface. That is why freeze isolation is not accepted in these plants and why a stalled actuator is treated as information rather than as a fault to be overcome with a bar.

How much heat tracing does a molten salt valve need?

For a DN200 body with 100 millimetres of insulation in still air, standing loss is roughly 350 watts at 565 C, which is about 9 metres of a 40 watt per metre cable plus the flanges and the stem. The number that matters more than the wattage is the setpoint. Holding the body at 500 C rather than 250 C keeps a cold start inside the 80 kelvin elastic budget of the wall, and costs only the standing loss during standby.

Can a standard Class 300 ball valve be used for molten salt?

The casting can, and almost nothing else can. At 20 bar and 565 C the pressure only asks for a 2.9 millimetre wall against the 13 millimetres the flange standard provides, so the body is the easy part. What fails is the seat material, the packing, the cavity relief in the reverse direction, the vent bore and the trace heating. A standard valve bought against a class and a size will be delivered with none of those, which is why the data sheet has to carry the salt, the band and the relief path rather than just the tag data.

Related Products

Explore the Valves in This Article
Nominal Size: 1 inch / DN25, range 1/4 to 12 inch Pressure Class: ASME Class 150, 19.6 bar at 38°C; Class 300 optional Body Material: ASTM B148 C95800 nickel aluminum bronze Body Type: Two piece, bolted Ball: C95800, SS316 hard faced optional Stem: C95800 or SS316, blowout proof Seat: PTFE, RPTFE, PPL, or reinforced PTFE Packing: PTFE or graphite combined packing End Connection: Threaded NPT / BSPT, flanged RF to ASME B16.5 optional Service Temperature: -29°C to +150°C standard, +200°C high temperature version Seawater Velocity Limit: About 6 m/s clean, 4.5 m/s with sand Design Standard: API 608, ASME B16.34 Test Standard: API 598, ISO 5208 Operation: Lever standard, pneumatic or electric optional
Size Range: 2 to 56 inch (DN50 to DN1400) Pressure Class: ASME Class 150 to 2500 / PN20 to PN420 Body Material: WCB, LCB, CF8, CF8M, CF3M, duplex, alloy steel Ball: Stainless with hard chrome, ENP, Stellite or tungsten carbide overlay Seat: PTFE, RPTFE, PPL, or metal Stellite faced Seat Arrangement: SPE, DPE (DIB-1), or combination (DIB-2) Packing: Live loaded graphite, ISO 15848-1 qualified Fire Test: API 607, also API 6FA and ISO 10497 Fugitive Emission: ISO 15848-1 class B or A, with temperature and cycle class stated Face to Face: ASME B16.10 short or long pattern End Connection: Flanged RF / RTJ to B16.5 and B16.47, butt weld to B16.25 Features: Double block and bleed, sealant injection, anti-static, anti-blowout stem Design Standard: API 6D, API 608, ASME B16.34, ISO 14313 Operation: Lever, gear, pneumatic, electric, hydraulic, gas over oil Sour Service: NACE MR0175 available
Nominal Size: 1-1/2 inch / DN40 Pressure Rating: 1000 WOG (68.9 bar cold working, non-shock) Body Type: Three piece, bolted end caps, four body bolts Body Material: ASTM A351 CF8 (304), CF8M (316) optional Ball: CF8, polished Ra 0.4 µm, full or reduced bore Seat: PTFE, RPTFE, PPL, or Stellite faced metal Body Gasket: PTFE to 180°C, graphite above Packing: PTFE V-ring to 200°C, graphite above End Connection: NPT, BSP, socket weld B16.11, butt weld B16.25, flanged RF Actuator Mounting: ISO 5211 F05 / F07 pad Air Supply: 0.4 to 0.7 MPa clean dry air Service Temperature: -29°C to +260°C, seat dependent Design Reference: MSS SP-110 general purpose, ASME B16.34 body wall Test Standard: API 598, ISO 5208
Nominal Size: 3/4 x 1/2 inch (DN20 x DN15), reduced bore Pressure Class: ASME Class 600 / PN100, 102.1 bar at 38°C Body Material: ASTM A182 F51 forged duplex, UNS S31803 PREN: Approximately 35 Ball: F51 with HVOF tungsten carbide or Stellite overlay Stem: F51, SS316 or 17-4PH, blowout proof Seat: RPTFE, Devlon, PEEK, or Stellite faced metal Bolting: B8M Class 2, alloy 718, or B7M for sour service End Connection: NPT, BSP, socket weld ASME B16.11, butt weld ASME B16.25 Bleed Port: Cavity bleed with plug or needle valve, lockable option Service Temperature: -29°C to +300°C, seat dependent Design Standard: API 6D, API 608, ASME B16.34 Test Standard: API 598, ISO 5208, cavity test through bleed port Sour Service: NACE MR0175 available