Lined Valves: Why Liners Fail When The Chemical Compatibility Chart Says They Are Fine

Almost every lined valve failure starts the same way. The process engineer checks the medium against a compatibility chart, the chart says PTFE is inert, the valve is purchased with a PTFE lining, and eighteen months later the plant is looking at a liner that has blistered away from the body, a bore that has been sucked flat, or a valve that holds 10 bar but weeps at half a bar. The chemistry was never wrong. The chart was simply answering a question nobody asked.

A compatibility chart tells you whether a polymer will be chemically attacked. It says nothing about whether small molecules will diffuse through the polymer, nothing about what happens to a polymer that is held under permanent compressive load at 150 °C, nothing about a material that grows eleven times more than the steel it is bonded to, and nothing about a bolted joint where the lining itself is the gasket. Those four mechanisms account for most of the failures in lined valves, and all four of them are invisible on a compatibility chart.

This article works through them in the order a specifier meets them: how the lining is made and why the process route sets the failure mode, which polymer belongs in which service, why the pressure rating quietly collapses above 100 °C, the permeation arithmetic and what it does behind the liner, vacuum collapse, cold flow in the seat, the expansion mismatch that cracks liners at the flange radius, and the flange joint that is where lined valves actually leak. Seat tightness acceptance is covered separately in our guide to ISO 5208 seat leakage classes, and the shell and seat test procedure is in the API 598 pressure testing guide; neither of them will tell you whether the lining will still be there in year three.

1. The Compatibility Chart Answers One of Four Questions

Before going further it is worth being precise about what a lining has to survive, because the four requirements are satisfied by four different properties and no single material is best at all four.

Chemical attack on the polymer. This is what the chart covers: does the medium dissolve, swell, oxidise or depolymerise the lining. For PTFE and PFA the answer is "no" for almost the entire periodic table, with a short and well-known list of exceptions: molten alkali metals, elemental fluorine, and a few aggressive fluorinating agents such as chlorine trifluoride. Solve this and you have solved the easiest of the four problems.

Permeation through the polymer. A liner is a barrier, not a wall. Small molecules - water vapour, hydrogen chloride, wet chlorine, hydrofluoric acid, ammonia, methanol - dissolve into the polymer at the wetted face, diffuse down the concentration gradient, and emerge on the back. The polymer is unaffected. The steel behind it is destroyed. Nothing on a compatibility chart addresses this, and it is the single most common cause of lined valve failure in hydrochloric acid, chlorine and HF service.

Mechanical load at temperature. A lining is a structural element. It carries bolt load at the flange, disc or ball load at the seat, and the full differential pressure across any unsupported span. Fluoropolymers creep, and they creep much faster above their glass transition, which for PTFE sits at 120 to 130 °C - below the 180 °C for which the same material is routinely sold. Chemical resistance does not decline at 130 °C. Everything mechanical does.

Differential expansion against steel. PTFE expands at 100 to 150 × 10^-6 per K. Carbon steel expands at about 12 × 10^-6 per K. The ratio is around eleven to one, and a liner cannot move eleven times more than the body it is locked into without developing stress somewhere.

A specification that names only the polymer grade is therefore controlling one of four variables. The rest of this article is about the other three.

2. How the Lining Is Made, and Why the Process Sets the Failure Mode

PTFE cannot be melt-processed. Its melt viscosity is around 10¹⁰ Pa·s - it does not flow even above its 327 °C melting point - so a PTFE liner is made by compressing powder into shape and sintering it, fusing the particles together under heat. PFA was developed to solve exactly this problem: it carries perfluoroalkoxy side groups that interrupt the chain enough to let it melt and flow, which means a PFA liner can be transfer-moulded or rotolined directly into the valve body as one continuous, dense, seamless skin.

The processing difference has three consequences that matter more than the datasheet's "chemical resistance: excellent" line for both materials.

Sintered PTFE keeps a small amount of porosity. Voids remain between the fused powder particles. They are invisible, they do not leak liquid, and they are the pathways that vapour permeates through. At equal thickness a melt-processed PFA lining typically shows 50 to 70 per cent lower permeation than a sintered PTFE one. That is a process fact, not a chemistry fact, and it is the real reason PFA is specified for chlorine and HCl.

Melt processing allows mechanical anchoring. A transfer-moulded PFA lining is forced into dovetail grooves or anchor holes machined into the casting, so the polymer is physically keyed into the metal rather than merely sitting against it. A sintered PTFE liner is usually a moulded sleeve, flared over the flange faces and held by the joint, with no anchor at all except friction and the flare. This is the difference between a lining that resists vacuum and one that does not.

The flare is the seal. On any lined valve, the lining is turned out over the flange face so the bolted joint closes polymer against polymer. Process fluid never touches steel - that is the whole point of the design. It also means the flange face is a 2 to 3 mm polymer washer carrying the entire bolt load, and that washer is the weakest structural element in the valve.

Three construction families follow from this, and each fails differently. A loose or flared sleeve is cheap, replaceable in principle, and the least resistant to vacuum and thermal cycling. A bonded or moulded-in lining with adhesive or dovetail anchoring resists vacuum but transfers thermal expansion stress into the bond line, which is where it cracks. A rotolined seamless body with encapsulated disc or ball is the premium route, eliminates seams and holidays, and still cannot escape permeation or creep - it just removes the joint as a failure site.

3. Which Polymer Belongs in Which Service

The fluoropolymer family is often treated as interchangeable because every member is "inert". They are not interchangeable on temperature, permeation, creep or abrasion, and the price difference between PTFE and PFA is large enough that the choice is a real engineering decision rather than an upsell.

Two rows deserve comment because they are where specifications usually go wrong.

The rubber rows are not a cheaper substitute for PTFE - they are the correct answer for a different question. Fluoropolymers have Shore D hardness in the 50 to 65 range, are soft, and have poor abrasion resistance. A PTFE-lined valve in an acid slurry with 10 per cent solids will lose its lining to erosion in months, at the disc edge and at the seat land, while a rubber-lined valve in the same line runs for years. Conversely, rubber swells in solvents and is attacked by oxidising acids, so it cannot simply be substituted back. Chemical resistance and wear resistance are separate properties, and a lined valve specification that names only one of them will be wrong about half the time.

The mPTFE row matters because it is the answer to a failure that looks like a manufacturing defect. Standard PTFE in a seat is under permanent compressive load from the disc or ball, and it takes a permanent set. Seat load decays, interference is lost, and the valve starts to weep - not at rated pressure, where the differential helps push the seal closed, but at low pressure where it does not. Modified PTFE contains a small amount of comonomer that reduces crystallinity and porosity, giving it meaningfully better recovery under sustained load while keeping PTFE's friction and chemical profile. If a lined valve has to hold bubble-tight at 0.5 bar after a thousand cycles, the seat should not be standard PTFE.

4. The Pressure Rating Collapses Above 100 °C

The most expensive misunderstanding in lined valves is that a PN16 valve is a 16 bar valve. It is a 16 bar valve at ambient, and the lining system - not the body - sets what happens above that. Above roughly 100 °C the permissible pressure falls, roughly linearly, to a fraction of the nominal rating at the top of the temperature range.

These are typical published envelopes and any individual manufacturer's curve governs, but the shape is universal and the mechanism behind it is worth understanding because it tells you which failures to expect.

Part of the derating is permeation: permeability rises roughly exponentially with temperature, so a hotter valve passes more aggressive vapour to its own backing. Part of it is creep: PTFE's glass transition at 120 to 130 °C is not a melting point, but it is where the amorphous phase stops being glassy and the creep rate under sustained load rises sharply. A seat that held its interference at 80 °C will take a set at 150 °C within a few hundred cycles. And part of it is simple strength: PTFE's Young's modulus of 400 to 750 MPa at room temperature is already two orders below steel, and it falls further as the material passes through the transition.

The practical consequence is a specification rule that catches a great many projects: the pressure class is a body property and the pressure rating at temperature is a lining-system property, and the second is always lower. A line at 8 bar and 170 °C in a PN16 PTFE-lined valve is outside the envelope even though both numbers individually look compliant, and no amount of body thickness fixes it. The remedies are to move to PFA, to move to a smaller temperature, to use an alloy valve with no lining at all, or to accept a lower pressure class and check the flange rating separately.

5. Permeation: Three Steps, One Equation, and a Temperature That Doubles It

Permeation is solution-diffusion, and it is worth writing down because the equation tells you immediately which levers exist and which do not. The steady-state flux through a liner is:

J = P × A × Δp / t

where P is the permeability of the polymer to that specific species, A is the wetted area, Δp is the partial pressure difference across the liner, and t is the liner thickness. Three of those four terms are fixed by the process and the valve size. Only thickness is available to the designer, and thickness is the weakest lever available, because permeability itself is strongly temperature-dependent:

P = P₀ × exp(-E_p / RT)

The activation energy for a small molecule diffusing through a fluoropolymer is typically in the region of 35 kJ/mol. Put that into the equation and a 40 K rise in service temperature multiplies the permeation rate by about 4.6. To recover the original flux by thickness alone, a 2.5 mm liner would have to become 11.5 mm - which is not a valve lining any more, it is a block of polymer that will not close.

The derivation is worth doing once because it settles an argument that recurs on every debottleneck. A plant wants to raise a hydrochloric acid transfer line from 40 °C to 80 °C. The lining is compliant at both temperatures and the compatibility chart is unchanged. But permeation rises 4.6-fold, and the entire flux that was previously diffusing into the annulus and escaping through the vents now arrives at four and a half times the rate. If the vents are undersized, partially blocked, or absent, the annulus pressure rises accordingly. The failure will be attributed to the medium, the supplier, or the lining material. It is none of those. It is a temperature change that moved the one variable the chart does not show.

This is also why "permeation rate" belongs in a lined valve datasheet and almost never appears there. The useful numbers are the polymer, the thickness, the vent arrangement, and the maximum service temperature - and the last of those is the one that actually governs.

6. What Permeation Does Behind the Lining

The permeate does not disappear when it reaches the back of the liner. It arrives in a thin annulus between polymer and steel, and what it does there determines whether the valve fails slowly or catastrophically.

It corrodes the backing. A carbon steel body behind a PTFE liner in wet HCl service is a corrosion cell with a continuous fresh supply of acid and no way to passivate. Iron chloride forms, its volume is several times that of the iron consumed, and the expansion pushes the liner away from the wall. The steel thins where it is needed most - at the flange and around the seat - and the liner loses the support that was carrying the differential pressure.

It builds pressure. If the annulus is sealed, permeated vapour accumulates and the liner is loaded from behind. The polymer bulges into the bore as a blister, usually at the hottest, thinnest or least supported point. On the next depressurisation the trapped gas expands and the blister can invert.

It needs somewhere to go. This is what a vented body is for. A tell-tale hole drilled through the steel backing to atmosphere turns an accumulating pressure into a slow, harmless release, and - more valuable than the release itself - it turns an invisible failure into a visible one. A vent that is wet, stained or fuming is a diagnostic, and it means the lining is passing something it should not. The single worst thing an operator can do to a vented lined valve is plug a weeping vent because it looks like a leak. It is the relief path, and blocking it guarantees the blister it was preventing.

The body vent is also the cheapest quality signal available at goods-in: a valve with no vent at all in a permeating service was built for a different application, regardless of what the datasheet says.

7. Vacuum, the Trapped Annulus, and What the Damage Looks Like

Lined valves fail in vacuum in a way that surprises people who assume a vacuum is a gentle condition. The load is only one bar, which is nothing compared to the 16 bar the body was tested at. The problem is not strength, it is stability, and it is not really the vacuum that causes it.

Start with the arithmetic. A DN 100 bore is 7.85 × 10^-3 m², so one bar of differential across it is 785 N. Spread across a 3 mm polymer disc spanning that bore, and treated as a clamped circular plate under uniform load, the peak bending stress is roughly 20 MPa. PTFE's tensile strength is only around 28 MPa at room temperature and considerably less at 150 °C, and its modulus is a fraction of a metal's, so the liner does not burst - it deflects, yields and takes a permanent set into the flow path. A DN 300 valve has nine times the area at the same stress, and a much larger unsupported span.

Now the part that is usually missed. Under vacuum inside the valve, the liner is pushed outward against the body, which supports it. That is not a failure. The failure comes from the annulus: if air, or accumulated permeate, is trapped between liner and body at atmospheric pressure and the bore is evacuated, the liner sees one bar from behind with nothing but its own stiffness and its anchors holding it. That is why a vented body helps in vacuum as well as in permeation - it equalises the annulus instead of letting it become a pressure source.

The resulting damage has a signature that is genuinely useful in a failure investigation, and it is documented in the phosgene safe-practice guidance published by the American Chemistry Council for lined piping, where the distinction is a safety-critical one. A liner that has failed by permeation collapse typically shows a single fold or lobe protruding into the bore. A liner that has failed by vacuum collapse typically shows three folds arranged around the bore. Open the valve, look at the shape of the deformation, and the mechanism tells you which specification clause was missing.

Which leads to the only reliable way to buy a vacuum-rated lined valve: ask for a number. "Vacuum rated" is a marketing phrase. The engineering question is at what absolute pressure, in what size, at what temperature, and for how many cycles - because the dovetail anchoring that resists full vacuum at 20 °C is a different proposition at 150 °C, where the polymer is softer and the anchor is carrying a load the modulus can no longer support. A supplier who cannot state the vacuum limit at temperature has not tested it.

8. Cold Flow: the Seat Is a Polymer Under Permanent Load

Lined valves leak for a reason that has nothing to do with wear, damage or chemistry: the seat relaxes. In a lined butterfly valve the liner is the seat, and it is compressed by the disc every time the valve closes and by the flange bolts permanently. In a lined ball or plug valve the seat ring is a polymer component under continuous preload. Fluoropolymers are viscoelastic, which is a polite way of saying that under a constant load they keep moving.

The mechanics are simple and the symptoms are specific. Seat interference is designed in - typically a few tenths of a millimetre of squeeze at the disc edge or on the ball. As the polymer creeps, that interference decays. The valve still closes, the handle still turns, the actuator still strokes, and the seat still holds when the differential pressure is high enough to push the disc or ball into the seat. At low differential it does not, because there is no longer enough stored elastic energy in the seat to follow the geometry. A lined valve that passes its seat test at 10 bar and weeps at 0.5 bar is not a valve with a bad test. It is a valve whose seat has taken a set, and the test result was correct at both pressures.

Three things make it worse, and all three are common. Temperature above the glass transition accelerates creep by a large factor, so a PTFE-lined valve rated to 180 °C will lose seat load far faster at 160 °C than at 60 °C even though both are within the rating. Thermal cycling resets the polymer's internal stress state each time and ratchets the set. And any design where the seat is loaded by bolt torque rather than by a spring has no way to recover the loss - the bolts relax, and nothing follows.

The countermeasures are material and mechanical, not procedural. Modified PTFE or TFM for the seat instead of standard PTFE. A spring-energised or pressure-assisted seat so that differential pressure supplements the preload rather than replacing it. Belleville washers under the body bolts so that bolt load is maintained as the joint settles. A seat thick enough that a few tenths of a millimetre of creep is a small fraction of the squeeze. And a realistic re-torque interval written into the maintenance plan, because a lined valve is one of the few items in a plant where the correct preventive maintenance is to tighten something on a schedule.

9. The Expansion Mismatch: Eleven to One

This is the mechanism that produces the crack nobody can explain, at the place nobody inspects - the radius where the flange flare turns into the bore.

Take a DN 200 lined valve. PTFE at 100 to 150 × 10^-6 per K and steel at 12 × 10^-6 per K, over a 200 mm span and a 150 K swing from ambient to 170 °C service. The PTFE wants to grow 3.0 to 4.5 mm. The steel grows 0.36 mm. The difference - roughly 2.6 to 4.1 mm, or 1.3 to 2 per cent of the span - has to go somewhere.

It cannot go into elastic strain: 1.5 per cent is far beyond what a fluoropolymer can hold elastically, and the polymer is the softer member. So it goes into one of three places. The liner buckles and wrinkles, which produces the wavy bore surface that later becomes a leak path at the seat. The liner extrudes, which thins the flare at the flange and eventually shears it. Or the liner slides, which means it "walks" out of its dovetail grooves over a few hundred cycles and ends up unsupported exactly where support was the design intent.

The crack, when it appears, is a radial split at the flare radius, and it is a fatigue crack rather than an overload one - which is why the valve passed its initial test perfectly and failed two years later. Each thermal cycle puts the flare through a strain cycle that the polymer accommodates by yielding. After enough cycles it separates.

Two operational habits drive this far harder than the design temperature does. Steam-in-place or CIP at 130 °C followed by a cold water flush is a 110 K shock in a few minutes, which is a much more severe event than slow heating to the same temperature. And repeated draining and refilling with cold medium on a hot valve does the same thing in reverse. If the process does this daily, the lining is being fatigue-tested daily, and the cycle count - not the calendar age - is what determines life.

The design answers are the same shape as the problem: give the liner somewhere to go. A controlled expansion detail at the flange, a flare long enough to flex instead of being pinched, a lining thickness that tolerates the strain, and a warm-up procedure that limits the rate to something the polymer can follow. A lined valve brought to temperature over 30 minutes lasts; the same valve brought to temperature in three minutes with a steam lance does not.

10. The Flange Joint Is Where Lined Valves Actually Leak

Ask a maintenance engineer where lined valves fail and the answer is rarely the bore. It is the joint. There are four reasons, and every one of them is a specification or assembly detail rather than a material property.

Flat face only. The lining is flared across the full flange face, so the mating flange must be flat faced - ASME B16.5 flat face, or EN 1092-1 type B. Bolting a lined valve to a raised-face flange concentrates the entire bolt load on the small raised annulus, which is precisely where the flare has to be supported, and it will crack or extrude the lining. This is not a preference. It is a compatibility requirement, and mismatched facings are one of the most common findings in a failed lined installation.

A full-face soft gasket, not a spiral wound one. The joint seals polymer to polymer. A spiral wound gasket needs a high seating stress to work, and that stress has to pass through the liner flare to get there. The correct gasket is a full-face soft gasket - PTFE envelope with an elastomer insert is the usual choice - sized to the bolt circle so the load is spread over the whole face rather than an annulus.

The bolt torque window is narrow in both directions. Under-torque and the joint leaks past the flare. Over-torque and the flare cold-flows: a 3 mm flare compressed hard enough will thin to half that, and once it is thin it creeps continuously and the joint loosens itself. The window is typically specified by the valve manufacturer as a torque figure per size, and it should be applied with a calibrated wrench in a star pattern, in at least two passes, because a single pass to final torque loads one side of the flare before the other has closed.

The joint is not reusable. A compressed fluoropolymer flare has taken a set. Dismantle the joint, and the flare will not return to its original thickness, so re-assembly at the same torque gives less load than the first time. Any lined valve that has been unbolted needs a new gasket and a check of the flare thickness before it goes back in.

The diagnostic value of all this is high: a leak that appears at the flange within days of installation is an assembly or facing problem; a leak that appears at the flange after two years of thermal cycling is a thinned flare. Both look identical from the outside, and the difference is whether the fix is a torque procedure or a new valve.

11. Failure Signatures and What Each One Tells You

Lined valve failures are usually diagnosed by guesswork because the evidence is inside a closed body. The table below maps what is actually found on strip-down to the mechanism that produced it, which in turn points at the clause that was missing from the specification.

The two rows that repay the most attention are the first two, because they are the ones that look like a material defect and are not. Both are annulus pressure problems, both are solved by a vent and by anchoring, and both are invisible in a quotation unless the buyer asks for them by name.

12. Velocity and Abrasion: the Property the Chart Does Not List

Fluoropolymer linings are soft. Shore D hardness of 50 to 65 is roughly the hardness of a hard hat, and the material has no abrasion resistance worth the name compared with rubber or with a metal seat. In clean chemical service this never matters. In slurry, crystallising or high-velocity service it is the dominant failure mode, and it is concentrated at two places.

The first is the disc edge of a lined butterfly valve. At part-open positions the flow accelerates past the disc edge and the liner is the surface it scrubs. The second is the seat land on any lined valve, where the clearance is smallest and the velocity is highest. Published velocity limits for lined butterfly valves are typically around 3 to 4.5 m/s for liquids, and they are not conservative suggestions - they are the point at which erosion of the lining starts to compete with chemical life. A valve sized for a 2 m/s line velocity but operated at 40 degrees open is seeing several times that at the disc edge, and that is where it will fail.

The correct answer for abrasive duty is usually not a thicker fluoropolymer but a different lining. Soft rubber for fine slurries and crystallising solutions, hard rubber or ebonite where the abrasion is continuous and the chemistry permits, and a metal-seated valve where neither polymer survives. This is a case where upgrading the specification - from PTFE to PFA, say - spends money on the wrong property entirely. If the failure is erosion, buying better chemical resistance changes nothing.

Velocity is also worth checking as a sizing question rather than a materials one. If the line is oversized, the valve runs further closed for the same flow, and the local velocity at the throttling element goes up even though the nominal line velocity looks comfortable. Our Cv and Kv sizing guide covers the arithmetic; the point here is that a lined valve's erosion life is set by the velocity at the seat, not by the velocity in the pipe.

13. Fire, Static and the Two Certificates Lined Valves Cannot Always Earn

Two common specification lines need challenging before they are written into a lined valve datasheet.

API 607 fire safety. A fluoropolymer lining is a polymer in the wetted path, and a fire test that burns the valve will consume it. A lined valve can only claim fire safety if it has a metal backup seat that closes the flow path after the lining is gone, and that is a specific design, not a property of the lining. Putting "API 607" in a lined valve specification without confirming the design is one of the more reliable ways to receive a certificate for a different product than the one being supplied.

Antistatic construction. Fluoropolymers are excellent insulators - volume resistivity above 10¹⁸ ohm·cm. In flammable service, a lining that cannot dissipate charge is a static accumulator, and the charge has to be grounded through the stem and body rather than through the polymer. Conductive grades of PTFE and PFA exist and are the usual answer for solvent and hydrocarbon duty, together with a verified grounding path. A standard lining plus a grounding strap on the flange is not equivalent, because the charge is generated at the wetted surface, not at the pipe.

Neither of these is exotic. Both are routinely omitted, and both are cheap to specify at the quotation stage and impossible to retrofit.

14. What to Put in the Purchase Specification

Most lined valve failures trace back to a specification that named a polymer and a pressure class and left the rest to the supplier. The clauses below are the ones that actually determine whether the valve survives, in rough order of how often their absence causes a failure.

A specification built from those twelve lines will not prevent permeation, creep or differential expansion - nothing prevents those. It will make sure the valve was designed for them, and it will make sure that when something does go wrong, the strip-down evidence points at a mechanism instead of at a supplier.

15. The Wider Point: Linings Are Mechanical Components

The reason lined valves disappoint is that they are bought as chemical products and installed as mechanical ones. The purchasing decision is made on a compatibility chart, which is a chemistry document. The valve then spends its life being compressed, heated, cooled, evacuated and eroded, which are mechanical and thermal loads, and none of them appear on the chart.

The four mechanisms in this article - permeation, annulus pressure, creep and expansion mismatch - explain most of the failures that get blamed on the lining material. Each has a specification answer that costs little at the quotation stage: a vent, an anchor, a modified polymer in the seat, a flat face and a torque table. Each is close to impossible to add after the valve is installed.

For buyers comparing quotations, the useful question is not "is the lining PTFE or PFA" but "show me the pressure-temperature curve for this size, the vacuum rating at temperature, the anchoring method, the spark test record, and the bolt torque table". A supplier who has all five is selling an engineered product. A supplier who has one is selling a polymer.

The range of lined and special-material valves we build for chemical, chlor-alkali, pharma and acid service is listed in the special material valve range, and where the medium is both corrosive and abrasive the comparison usually comes down to a lined butterfly valve against a lined ball valve on shut-off class, cavity behaviour and torque rather than on lining material. Where the metallurgy is the answer instead of the lining - chloride service above the practical temperature of any polymer - the selection logic in our duplex and super duplex guide applies, and the two routes should be compared on total cost rather than on unit price.

16. Frequently Asked Questions

Is PFA always better than PTFE for a lined valve? 

No. Both resist essentially the same chemicals, so the choice is about the three properties that are not chemistry. PFA is melt-processed, denser and anchored in dovetail grooves, so it permeates less, holds under vacuum and tolerates thermal cycling better; it costs more, and the price gap grows with size. Standard PTFE is the economical answer for stable, non-permeating acid and alkali duty in large sizes. The rules of thumb that hold up in practice are: PFA for hydrochloric acid, wet chlorine, bromine, HF and any vacuum or cycling duty; PTFE for neutral salts, dilute acids at moderate temperature and water treatment; and modified PTFE (mPTFE or TFM) for the seat either way, because the seat fails by creep rather than by chemistry.

Can a lined valve be used under full vacuum? 

Only if the supplier states a vacuum rating in mbar absolute, for that size, at that temperature - and only if the body is vented. The load is small in absolute terms: one bar across a DN 100 bore is about 785 N, which a 3 mm polymer disc carries at roughly 20 MPa of bending stress, below its short-term strength but well above what it can hold without taking a permanent set once it is hot. The real mechanism is the annulus: if air or accumulated permeate is trapped between liner and body and the bore is evacuated, the liner is pushed into the flow path from behind. Dovetail anchoring plus an open vent is what makes a lining vacuum-capable; a flared sleeve with no anchor is not, whatever the datasheet claims.

Why does a lined valve pass its pressure test and then leak in service? 

Because the two tests are measuring different things. A hydrostatic seat test at rated pressure loads the seat with the differential pressure itself, which pushes the disc or ball into the seat and assists the seal. At low differential pressure there is no such assistance and the seal depends entirely on stored elastic energy in the polymer - which decays as the seat creeps. A lined valve that is tight at 10 bar and weeps at 0.5 bar is reporting lost seat interference, not a faulty test. The same logic applies to the body: a shell test at 1.5 times the rating proves the casting, and says nothing about whether the lining is still bonded to it.

What flange facing and gasket does a lined valve need? 

Flat face mating flanges and a full-face soft gasket. The lining is flared across the whole flange face so that the joint seals polymer to polymer and the medium never touches steel; a raised-face flange concentrates the bolt load on the small raised annulus exactly where the flare needs support, and will crack or extrude it. Do not use a spiral wound gasket - it needs a high seating stress that has to pass through the flare to be developed. Torque to the manufacturer's table per size, in a star pattern over at least two passes, and treat the joint as non-reusable: a compressed flare has taken a set and will not seal twice at the same torque.

How do I tell whether a liner failed from permeation or from vacuum? 

By the shape of the deformation. A liner that collapsed because permeate built pressure behind it typically shows a single fold or lobe protruding into the bore; a liner that collapsed under vacuum with a trapped or blocked annulus typically shows three folds arranged around the bore. The distinction is used in the phosgene safe-practice guidance published by the American Chemistry Council for lined piping, and it is worth using in valve failure reports too, because the two failures have different fixes: one needs a lower service temperature or a denser lining, the other needs an open vent and mechanical anchoring.