Views: 0 Author: J-VALVES Publish Time: 2026-10-19 Origin: Site
Start with the classification, because most of the confusion in this product category is a documentation problem rather than an engineering one.
ISO 15848-1:2015+A1:2017 classifies a valve on three axes at once: tightness class, endurance class, temperature class. The tightness limits are measured either as helium mass flow per metre of stem perimeter or as methane concentration:
Tightness class | Methane | Helium | Typical construction that reaches it |
A | ≤ 50 ppmv | ≤ 1 × 10⁻⁵ mg/(s·m) | Bellows, or an equivalent rotary stem seal |
B | ≤ 100 ppmv | ≤ 1 × 10⁻⁴ mg/(s·m) | PTFE or elastomer stem seal |
C | ≤ 500 ppmv | ≤ 1 × 10⁻² mg/(s·m) | Flexible graphite packing |
The body and cover joints are held to ≤ 50 ppmv regardless of stem class, which is worth remembering: a valve can pass Class A at the stem and still fail on the bonnet gasket.
Endurance is where the numbers get misquoted. There are two separate scales, and they differ by two orders of magnitude:
Endurance class | Mechanical cycles | Thermal cycles | Stem seal adjustments permitted |
CO1 (isolating) | 205 | 2 | 1 |
CO2 (isolating) | 1,500 | 3 | 2 |
CO3 (isolating) | 2,500 | 4 | 3 |
CC1 (control) | 20,000 | 2 | 1 |
CC2 (control) | 60,000 | 3 | 2 |
CC3 (control) | 100,000 | 4 | 3 |
A bellows seal globe valve used for on/off service is an isolating valve, so it belongs on the CO scale: 205, 1,500 or 2,500 cycles. Writing "ISO 15848-1 CC3" on the datasheet of an on/off globe valve is a common and expensive error — it advertises a 100,000-cycle endurance requirement that the product was never tested to, and it usually gets discovered during third-party review.
Temperature classes are −196 °C, −46 °C, −29 °C, ambient, 200 °C and 400 °C. So the only statement that means anything is a triple, for example AH / CO2 / t200. This is the single most quotable line in the article, and it is worth stating flatly:
An ISO 15848-1 rating is always a three-part designation. A standalone "Class A" is not a comparable performance claim — it does not say after how many cycles, or at what temperature, the class was still held.
One more trap sits next to this one. ISO 15848-2 is a production acceptance test, not a type test. It runs five mechanical cycles, at ambient temperature, with no thermal cycling, and its limits are A ≤ 50, B ≤ 100, C ≤ 200 ppmv — and the standard says explicitly that Part 2 classes are not equivalent to Part 1 classes. A factory acceptance report to Part 2 cannot be used to substantiate a Part 1 Class A claim. Reviewers who accept that substitution are accepting a document that says something different from what they think it says.
It helps to convert classes into mass, because concentrations hide the scale. Take a 25 mm stem, perimeter 0.0785 m, and annualise the helium-equivalent rate:
Stem seal condition | Helium-equivalent rate | Per valve, per year | 500 valves, per year |
AH (bellows) | 1 × 10⁻⁵ mg/(s·m) | 0.025 g | 12 g |
BH (PTFE / elastomer) | 1 × 10⁻⁴ mg/(s·m) | 0.25 g | 124 g |
CH (flexible graphite) | 1 × 10⁻² mg/(s·m) | 25 g | 12 kg |
Unmaintained graphite packing (field experience) | ~1 × 10⁻¹ mg/(s·m) | ~250 g | ~124 kg |
The table is useful because it reframes a class as a cost. AH to CH is three orders of magnitude; CH to a neglected packed valve is one. That is the arithmetic behind a common and often correct decision: fit bellows to the twenty or thirty valves where a stem leak is genuinely intolerable, rather than upgrading a thousand valves from C to B. The reductions live in the orders of magnitude, and the capital lives in the small population. Values are helium-equivalent and intended for magnitude comparison only, not as a mass flow of any specific process fluid.
"Design life: 10 years" is not a specification. A bellows counts strokes. Convert, and a surprising number of projects fail immediately.
Actuations per day | 10-year total | Against a 10,000-cycle budget | Verdict |
0.5 (3–4 per week) | 1,825 | 18 % | Bellows is not the limiting item |
2 | 7,300 | 73 % | Standard design covers it |
5 | 18,250 | 183 % | High-cycle bellows or oversized assembly |
10 | 36,500 | 365 % | Requires high-cycle design; expect cracking inside two years |
Continuous throttling | > 10⁶ | — | Bellows concept does not apply |
Stroke amplitude is the second consumer, and it is not linear. Bellows fatigue is low-cycle fatigue; in the plastic-dominated regime the Coffin–Manson relation gives life approximately inversely proportional to the square of strain amplitude (c ≈ −0.5). The practical consequence:
Halving the stroke amplitude roughly quadruples the cycle life; doubling it cuts life to about a quarter.
This produces a counter-intuitive field result. A valve that is "barely used" but is nudged through 5 % strokes several times a day accumulates damage faster than a valve that is stroked fully twice a week — the cycle count is high, and the strain per cycle never becomes small enough to ignore.
Temperature takes its cut from two directions. First, the fatigue strength of 316L at 350 °C is below its room-temperature value, and a 0.1–0.3 mm wall oxidises and thins with no corrosion allowance at all. Second, thermal growth eats the stroke budget directly. A 300 mm 316 stem rising from ambient to 350 °C, with α = 17.3 × 10⁻⁶ /K, grows by
300 × 17.3 × 10⁻⁶ × 320 ≈ 1.66 mm
That 1.66 mm does not disappear. The bellows absorbs it — in compression if the valve is closed, in extension if it is open. If the bellows was sized for the valve stroke alone, the valve is over-compressed for its entire hot life. This is one of the most common causes of a "new valve cracked at the crest in three months" report.
Stroke budget item | Value (worked example) | Basis |
Disc design stroke | 20.0 mm | Seat and disc geometry |
Thermal growth at 350 °C | +1.7 mm | 300 mm stem, ΔT = 320 K |
Manufacturing and assembly tolerance | +1.0 mm | Stem, bonnet, bellows stack |
Margin (20 %) | +4.5 mm | Prevents end-of-stroke compaction |
Required bellows design stroke | ≥ 27 mm | Convolution count and height follow from this |
The industry rule of thumb is that axial stroke should not exceed roughly 25 % of the free convolution length. Past that ratio the life curve falls steeply, and no amount of alloy upgrading recovers it.
The geometry of a bellows contains a dead zone by design. Between two adjacent convolutions there is an annular groove with a high depth-to-width ratio; the bulk flow does not reach it, and local velocity is close to zero. For clean media this is irrelevant. For media that crystallise, polymerise or drop solids, the groove is a trap.
The consequence is not "blockage." It is worse and quieter: the stroke is shared among all convolutions, so when some valleys fill, those convolutions stop flexing and the remaining ones must absorb the entire stroke. Peak strain rises, and because life scales with the inverse square of strain, the penalty is steep.
Convolutions immobilised | Strain per active convolution | Remaining life (estimate) |
0 % | 1.00× | 100 % |
10 % | 1.11× | 81 % |
25 % | 1.33× | 56 % |
50 % | 2.00× | 25 % |
A 20-convolution bellows with five valleys filled has lost roughly half its life while the valve still strokes smoothly, still passes a shell test, and still looks correct from the outside. There is no external symptom until the crack appears — which is why jacketing or flushing is not optional in crystallising service.
Three families of media cause this:
Crystallising: 32 % and 50 % caustic soda, molten urea, ammonium sulphate, brine. The material comes out of solution as soon as the bonnet cools, and repeated stroking compacts the crystals into a hard shell.
Polymerising: styrene, acrylates, dicyclopentadiene, isocyanates. Retention in the valley initiates local polymerisation; the product cannot be flushed out and the bellows assembly is scrap.
Solids-bearing: catalyst fines, slurry, pipe scale. Particles do not dissolve; they wedge into the valleys and act as abrasive.
The design response is a package, not a single feature: steam jacketing or electric tracing holding the bonnet 20–30 °C above the crystallisation point; a flush connection into the bellows cavity; vertical installation with the stem up so gravity drains rather than pools; and a written procedure for one full-stroke exercise per quarter. Full strokes crush and expel the deposit. Small modulating nudges compact it.
Phase belongs to the same limit. Under vacuum the bellows sees external pressure, and a standard axial bellows will squirm — the convolutions translate laterally, the pitch becomes uneven, and cracking follows at the flattened section. Steam systems produce the same loading without anyone specifying vacuum: close the valve, the steam condenses, and a negative pressure forms within minutes. If the datasheet never said "vacuum," the bellows was probably checked for internal pressure only.
The first two limits decide how much strain the bellows must absorb. The third decides how that strain enters it.
Anti-rotation comes first. A bellows has almost no torsional stiffness. It is designed to extend and compress along the stem axis. If the stem rotates with the handwheel, every convolution is twisted, and the resulting shear stress superimposes on the axial stress. The crack then runs at roughly 45° to the axis in a helix — a signature completely different from the circumferential fatigue crack, and the fastest way to tell a design fault from an assembly fault in the field. The correct arrangement drives a stem nut in the yoke and constrains the stem with a pin, key or guide so it translates without rotating.
The field check is simple: at the fully open position, apply handwheel torque and watch the stem. It should not rotate perceptibly. Measure the anti-rotation pin clearance; beyond about 0.5 mm the pin is no longer preventing rotation, it is merely limiting it.
Pressure thrust is second, and it is the economic ceiling on size. The bellows encloses the fluid, so pressure acts on its effective area and produces an axial force that must be carried through stem, yoke and stem nut:
Size | Approx. mean bellows diameter | Effective area | Thrust at Class 300 (5.1 MPa) | Thrust at Class 600 (10.2 MPa) |
DN50 | 78 mm | 0.0048 m² | 24 kN | 49 kN |
DN100 | 130 mm | 0.0133 m² | 68 kN | 135 kN |
DN150 | 185 mm | 0.0269 m² | 137 kN | 274 kN |
DN200 | 240 mm | 0.0452 m² | 231 kN | 461 kN |
These are upper-bound estimates before the seat-side reaction is subtracted; good designs bring the mean bellows diameter close to the seat bore so that most of the thrust cancels. Even so, above roughly DN150 at Class 600 the actuator demand and handwheel torque leave the economical range. This, not manufacturing difficulty, is why bellows sealing thins out above DN200 and is rare above Class 900.
The shell test is third, and it is the one that gets valves quietly damaged in the shop. A bellows valve must be shell-tested in the open, or at least partly open, position so that pressure acts on both faces of the bellows. Test it closed — the habit for every other valve type — and the bellows carries the full 1.5 × rating across a 0.1–0.3 mm wall on its own. The test passes, the valve is written off, and nobody records it. Put the requirement in the test procedure; do not leave it to shop practice.
The secondary seal is fourth. A bellows valve has more than one barrier: the bellows as primary, then graphite packing and a backseat outside it. The intent is that a cracked bellows hands sealing duty to the packing and buys time to plan a replacement. That intent has a precondition that is routinely skipped — the packing must actually be preloaded. Many manufacturers leave it loose precisely because the bellows is doing the work, which makes the backup nominal. When the bellows finally cracks, the packing behaves like an ordinary packed stem and leaks at an ordinary packed-stem rate.
Write the requirement into the purchase order: the secondary packing shall seal independently at full differential pressure to at least CM class (≤ 500 ppmv) after bellows failure.
Pair it with a tell-tale. A 1/4″ NPT port into the cavity between bellows and packing, connected to a pressure gauge or a gas detector, converts a completely hidden failure into an observable one. It is the cheapest and highest-return item in the whole specification.
# | Chain | Trigger | Signature | Field check | Design fix |
F1 | Over-stroke → circumferential crack at crest | Actuator limit mis-set, no mechanical stop, thermal growth excluded from stroke | Circumferential crack on the crest, fatigue striations | Measured stroke exceeds drawing; tell-tale pressurised | Mechanical stop, stroke calibration, thermal growth in budget |
F2 | Torsion → 45° helical crack | Worn anti-rotation pin, excessive key clearance, stem turned during assembly | Crack at ~45° to axis, crossing several convolutions | Torque at full open; stem visibly rotates | Pin or key plus periodic clearance check, ≤ 0.5 mm |
F3 | Valley deposit → local over-strain | Crystallising, polymerising or solids-bearing media; frequent short strokes | Filled valleys, uneven pitch, local flattening | Strip and inspect valleys for hard deposit | Jacket or trace, flush port, vertical mounting, scheduled full strokes |
F4 | Pitting or SCC → pinhole | Chlorides above threshold, wet chlorine, HCl, stagnant liquid | Single pinhole, no fatigue pattern around it | Wall thickness check, borescope | Upgrade to Inconel 625 or C-276; control chloride and stagnation |
F5 | End-plate weld failure | Weld defect, thermal cycling, assembly stress | Crack initiating at the circumferential weld | Helium leak test, dye penetrant | 100 % helium leak test, qualified WPS, replace as an assembly |
F6 | External-pressure collapse (squirm) | Vacuum service, steam condensation | Flattened convolutions, disordered pitch | Visual deformation, abnormal stroke | Design for external pressure, internal support, vacuum breaker |
The field triage is more straightforward than it looks. Read the crack direction first: circumferential means axial fatigue or over-stroke; 45° helix means torsion; a single pinhole means corrosion; a weld-root crack means fabrication; flattened convolutions mean external pressure. Get the direction right, then confirm with stroke records, pin clearance and a fluid analysis.
Bellows sealing is not a universal answer. In the following cases the correct move is to change the valve type, not to buy a more expensive bellows.
1. Continuous throttling or frequent modulation. The bellows is designed around full-stroke cycling. Its life model has no entry for three thousand 5 % nudges a day. A globe valve's trim is genuinely good at throttling, which puts the trim's strength in direct conflict with the bellows' fatigue logic. If the process needs modulation, specify a control valve — globe or cage — and treat modulation frequency as a life-reduction input rather than buying an on/off valve and using it as a regulator.
2. Solids, crystallisation or polymerisation. Unless jacketing, a flush connection and scheduled full-stroke exercise are all present, valley filling is only a matter of time. Slurry service belongs to full-port ball valves, knife gate valves or pinch valves, where "can it pass" and "can it seal" are solved separately.
3. Large size combined with high class. Above DN200 at Class 900 and up, thrust, bellows stiffness and actuator size amplify each other; cost leaves the acceptable range while reliability does not improve. That territory belongs to fully welded ball valves or metal-seated butterfly valves.
4. A requirement for on-line maintenance or live repacking. A bellows assembly is not a field-repairable part. It is normally welded into a bonnet sub-assembly, and replacement means shutdown, bonnet removal, whole-assembly swap and a new pressure test. Reading "less packing maintenance" as "maintenance-free" is the most common purchasing error in this category.
5. Instrument piping, and non-metallic or low-melting bellows. MSS SP-117-2023 explicitly excludes non-metallic bellows and bellows made from low-melting materials such as brass or phosphor bronze, and its scope is limited to steel globe and gate valves otherwise complying with API 600, API 602, API 603 and ASME B16.34. It does not cover ball, butterfly or check valves — for those, there is no equivalent component-level bellows standard to cite.
Safety limit, stated separately: the bellows is not an overpressure device. It withstands design differential pressure only and provides no relief function. Cavity overpressure from thermal expansion must be handled by self-relieving seats or an external relief valve. The bellows will not "give a little."
Material | Typical range | Suitable media | Limits |
316L | −196 to +400 °C | Water, steam, light hydrocarbons, general chemicals | SCC risk at Cl⁻ > 50–100 mg/L with T > 60 °C; no corrosion allowance |
321 / 316Ti | −196 to +500 °C | High-temperature steam, hot oil | Elevated-temperature strength still limited; demand a temperature-corrected life curve |
AM350 | −40 to +425 °C | Stem bellows needing high fatigue strength | Corrosion resistance below austenitics; not for aggressive media |
Inconel 625 | −196 to +600 °C | Hot chlorine, hydrogen, high-temperature flue gas | Cost; still verify the corrosion chart at actual concentration |
Hastelloy C-276 | −196 to +400 °C | Wet chlorine, hydrochloric acid, mixed acids | Highest cost; demanding fabrication and welding |
The governing principle is that a bellows has no corrosion allowance. A body can be designed with 3 mm of sacrificial thickness. The bellows has 0.1–0.3 mm, so the same corrosion rate that thins a body perforates a bellows. Corrosion rates must be checked against the bellows alloy specifically; extrapolating from body-material data is how pinholes appear in year three.
# | Clause | Wording | Reference |
1 | Emissions rating | ISO 15848-1 triple: AH (or AM) + CO2 + t200 | ISO 15848-1:2015+A1:2017 |
2 | Secondary seal | Packing shall seal independently to ≥ CM after bellows failure | Manufacturer test report |
3 | Bellows assembly | MSS SP-117-2023 compliant, 100 % helium leak tested | MSS SP-117-2023 |
4 | Cycle life | ≥ 10,000 full strokes, with test report stating stroke amplitude | Type test |
5 | Stroke budget | Vendor to calculate thermal growth and tolerances; design stroke ≥ calculated value | §5.2 |
6 | Anti-rotation | Pin or key design, clearance ≤ 0.5 mm, no stem rotation under full-open torque | General arrangement drawing |
7 | Tell-tale | 1/4″ NPT port between bellows and packing | General arrangement drawing |
8 | Pressure boundary | API 602 (NPS ≤ 4) or ASME B16.34; shell 1.5×, seat 1.1× | API 602 / B16.34 / API 598 |
9 | Test condition | Shell test with valve open; closed-position testing prohibited | Test procedure |
10 | Materials | Alloy, ply count, ply thickness and total wall stated on drawing | Quality dossier |
11 | Replacement strategy | Bellows assembly identified as a whole-unit replacement part, with part number and lead time | Maintenance manual |
12 | Exclusions | Vendor to confirm in writing: no solids, no crystallisation, not vacuum (or designed for external pressure) | Signed datasheet |
If the medium is chlorine, phosgene, HCN, acrylonitrile or similar acute toxins, then require bellows + secondary packing to CM + tell-tale to alarm. All three, not two.
If actuation is under 50 full strokes a year, then bellows life is not the constraint; a packed valve with live-loaded packing is adequate unless regulation says otherwise.
If the process modulates several times a week, then do not use a bellows globe valve; use a control valve. The bellows life model does not support modulation.
If solids exceed 1 % or the medium crystallises, then jacketing + flush port + vertical mounting + quarterly full strokes must all be present. If one is missing, change the valve type.
If design temperature exceeds 350 °C, then move from 316L to 321 or Inconel 625 and demand the life curve at that temperature, not at ambient.
If vacuum or steam-condensation negative pressure is possible, then the datasheet must state external pressure and the bellows must be checked for it; a standard internal-pressure design may collapse.
If DN ≥ 150 at Class ≥ 900, then estimate p × A_eff before choosing the operator. If thrust leaves the economic range, switch to a fully welded ball valve.
If the tender only says "complies with MSS SP-117," then add the ISO 15848-1 triple. SP-117 governs the bellows assembly; it says nothing about whole-valve fugitive emissions.
If a shell test is required, then run it open. Testing closed puts 1.5 × rating across the bellows alone — the test passes and the life is spent.
If the plant has bellows valves sitting permanently half-open, then inspect those first at the next outage — valleys and pitch uniformity, not leakage.
Is a bellows seal globe valve genuinely zero-leakage?
While the bellows is intact, the stem leakage path is structurally eliminated, so helium emissions can reach ISO 15848-1 class AH (≤ 1 × 10⁻⁵ mg/(s·m)) or methane class AM (≤ 50 ppmv). That applies only while the bellows is intact and the body and bonnet joints are sound. Once the bellows cracks from fatigue, corrosion or over-stroke, sealing duty passes to the secondary packing and the leak rate returns to packed-valve levels. The accurate description is near-zero leakage with a finite cycle limit — not permanent zero leakage.
If the bellows seals, why keep the packing at all?
Because a bellows is a fatigue component with a real probability of failure, and toxic service does not tolerate single-point failure to atmosphere. Packing and backseat form the second barrier, whose job is to take over after a crack and buy time for a planned replacement. The backup only works if the packing is properly preloaded. Left loose — which happens when the bellows is doing all the work — it will not seal when it is finally needed. Specify independent sealing to at least CM class after bellows failure.
How many years will it last?
Years is the wrong unit; cycles is the right one. The usual commercial commitment is ≥ 10,000 full strokes: about 13 years at two actuations a day, under three years at ten. Then apply the modifiers — halving stroke roughly quadruples life, higher temperature reduces it, and partial valley filling can cut the remainder to about half. The useful question is not "how many years" but "how many cycles, at what stroke and what temperature."
Can it be used as a control valve?
Not advised. Bellows life is built around full-stroke cycling; frequent small modulations concentrate strain in a few convolutions and accumulate damage faster than full-stroke operation, and higher velocity in modulating service accelerates valley deposition. The trim can throttle — that is a property of the trim, not of the bellows. Where modulation is required, use a control valve, or accept a documented life reduction and stock accordingly.
Is "ISO 15848-1 Class A" the same as "TA-Luft compliant"?
No. ISO 15848-1 is a classification system, and its classes are meaningless without endurance and temperature. TA-Luft is a national regulatory requirement that uses comparable measurement methods but states a compliance criterion rather than offering optional classes. The two cannot substitute for each other, and neither can be evidenced by an ISO 15848-2 factory acceptance report — Part 2 runs five mechanical cycles at ambient with no thermal cycling, and the standard states that its classes are not equivalent to Part 1.
What happens when the bellows cracks, and how would we know?
Fluid enters the cavity between bellows and packing. With a tell-tale port fitted, a gauge or detector responds immediately — the best available outcome. Without one, the packing either contains the fluid or passes it slowly, producing an irregular leak around the bonnet that is hard to distinguish from a packing problem. That is why the tell-tale is not an option but the mechanism that turns a hidden failure into an observable one. Once identified, take the valve out of service and replace the bellows assembly as a unit; replacing only the bellows usually fails because the old bonnet sealing face is no longer flat enough for a new assembly.
