When a flange weeps three days after start-up, the first conversation is always about the gasket. It is the wrong conversation. On a properly designed joint the gasket is the cheapest, softest and most forgiving element in the assembly, and it is usually doing exactly what it was asked to do. What failed was the preload, or the scatter in the preload, or the preload that was correct at 20 degrees C and no longer correct at 350, or a bolt that was hard enough to be strong and therefore hard enough to crack.
Bolting is where a valve stops being a product and becomes a joint, and it is the part of the supply chain with the least engineering attention per dollar spent. A DN200 Class 600 valve might carry twelve studs and twenty four nuts, all of them purchased on a line item that says "bolting: per flange", and all of them capable of taking the joint out of service. This article works through the arithmetic that decides the joint, the grades that are available to meet it, and the failure mechanisms that the arithmetic does not capture.
1. The joint is a spring, and the gasket is the soft part
A bolted flange joint is a preloaded spring system in parallel: the bolt acts as one spring, the flange and the gasket together as another, and tightening stores elastic energy in both. When pressure arrives it tries to push the flanges apart, and the joint survives because the bolts stretch a little more and the gasket is unloaded by only a fraction of what the pressure would otherwise take. That is the whole principle, and every rule that follows is about keeping the two springs in balance.
ASME VIII Division 1 Appendix 2 formalises it with two load cases. The seating case, Wm1, is what is needed to seat the gasket before any pressure exists: the gasket contact area times the seating stress y. The operating case, Wm2, is what is needed to hold the joint tight while pressure is applied, and it has two terms: the pressure end load trying to separate the flanges, and the residual gasket load that keeps the joint tight, which is the contact area times the gasket factor m times the pressure.
The magnitudes are instructive. Take a DN200 Class 600 joint at 102 bar with a spiral wound stainless and graphite gasket, an effective gasket width of about 8.4 millimetres and a gasket diameter of about 268 millimetres. The seating load works out near 0.49 meganewtons. The operating load is about 0.79 meganewtons, and of that, roughly 0.58 meganewtons is the pressure end load, about 73 percent, while the gasket term is the remaining quarter. In other words, at high pressure the joint is mostly a pressure vessel closure, and the gasket is a minority shareholder.
Quantity | Value for the DN200 Class 600 example | What drives it | Where it goes wrong |
|---|---|---|---|
Design pressure | 102 bar at ambient, Group 1.1 material | ASME B16.34 rating table | Using the 19 bar of a Class 150 as if it were a rule of thumb |
Gasket seating load Wm1 | about 0.49 MN | Contact area times y, 69 MPa for spiral wound | Under-torquing because the joint "looks tight" |
Pressure end load | about 0.58 MN | Bore area times pressure | Forgetting that it rises with pressure squared of diameter |
Operating gasket load | about 0.22 MN | Contact area times m times pressure, m of 3.0 | Reusing a gasket that has already taken a set |
Total operating load Wm2 | about 0.79 MN | Sum of the two above | Assuming the seating case governs at high pressure, it does not |
Bolt area required at 60 percent of B7 yield | about 1825 square millimetres | Wm2 divided by 0.6 times 724 MPa | Selecting bolts by hole size rather than by load |
2. Torque is an estimate, not a measurement
The torque wrench is the most trusted instrument on site and the least accurate. The relationship is T equals K times F times d, where T is torque, F is the desired preload, d is the nominal diameter and K is a friction coefficient that lumps together thread friction and under-head friction. That single number, K, is where the uncertainty lives: 0.20 for a dry as-received bolt, around 0.16 with a proper molybdenum disulphide lubricant, 0.12 with a heavy moly paste, and anything from 0.10 to 0.25 in the real world depending on coating, surface finish, temperature, re-use and how the lubricant was applied.
The consequence is that torque control delivers a preload with a scatter of plus or minus 25 to 30 percent even when everything is done correctly, and considerably worse when a bolt is lubricated by hand with an unknown product. Because friction consumes most of the torque, most of the energy going into the wrench never becomes preload, and the fraction changes with a coat of paint or a spot of rust inside a nut.
Three ways out exist, and they cost progressively more. Torque with a controlled lubricant, one product, applied by weight or by dipping, and a calibrated wrench, is the baseline. Torque plus turn-of-nut, where the nut is run down to snug and then rotated a defined angle so that the bolt is stretched by a known amount regardless of friction, is more accurate and needs a marking pen. Hydraulic tensioning, where the stud is stretched directly by a puller and the nut is run down by hand against the stretched length, removes friction from the calculation almost entirely, and it is the normal choice for Class 900 and above on large valves, for hydrogen service, and wherever a joint cannot be re-tightened later.
3. Bolting grades and what each one is for
The default for a carbon steel valve flange is ASTM A193 Grade B7, a chromium molybdenum steel, quenched and tempered to a minimum yield of 105 ksi, about 724 MPa, with a tensile strength of 125 ksi. It covers ambient and moderately elevated service, it is cheap and universally available, and it is the wrong choice in three situations that are easy to identify.
The first is sour service. NACE MR0175 limits carbon steel bolting to a maximum hardness of 22 HRC, which B7 does not guarantee, so the grade called for is B7M: same chemistry, deliberately under-strengthened to a minimum yield of 80 ksi and a maximum hardness of 22 HRC, with 100 percent hardness testing. The second is low temperature, below about minus 29 degrees C, where the impact requirements change and the grade becomes A320 L7, with L7M for the sour low temperature combination. The third is high temperature, above roughly 450 degrees C, where carbon and low alloy bolting loses creep strength and the answer is usually a precipitation hardened austenitic such as A453 Grade 660, or a nickel alloy for the top of the range.
Stainless bolting, A193 B8 and B8M, is common on stainless and duplex valves for appearance and for galvanic compatibility, and it is a trap if it is supplied in the annealed condition. Annealed 316 bolts have a yield around 200 to 210 MPa, which is a third of B7, and they will relax and lose preload on a joint that was designed around a high strength stud. The grade that is usually wanted is B8M Class 2, strain hardened, with a yield in the 550 to 700 MPa range.
Grade | Typical min yield | Temperature window | Hardness limit | Where it is the right answer |
|---|---|---|---|---|
A193 B7 | 724 MPa, 105 ksi | minus 29 to about 450 degrees C | Typically 35 HRC max | General carbon and stainless valve flanges, the default |
A193 B7M | 550 MPa, 80 ksi | minus 29 to about 400 degrees C | 22 HRC max, 100 percent tested | Sour service to NACE MR0175, wet H2S |
A320 L7 | 724 MPa, 105 ksi | down to minus 101 degrees C | 35 HRC typical | Cryogenic and low temperature, LNG and LPG |
A193 B8M Class 2 | 550 to 700 MPa | minus 196 to about 500 degrees C | No NACE approval at this strength | Stainless and duplex valves, marine atmosphere |
A453 Grade 660 | about 725 MPa at temperature | up to about 540 degrees C and above | Grade specific | Steam, hot oil, high temperature headers |
4. Hardness is the hydrogen switch
Hydrogen embrittlement is not a corrosion problem that happens to a bolt; it is a delayed fracture mechanism that starts with atomic hydrogen and ends with a crack, sometimes days after the joint was tightened and sometimes at 40 percent of the calculated tensile capacity. The hydrogen comes from three places: from cathodic protection or from galvanic corrosion at the bolt surface, from the process itself in hydrogen, wet H2S or ammonia service, and from the plating process if the fastener was electroplated.
The controlling variable is hardness, because hardness is a proxy for the strength and for the density of traps where hydrogen accumulates. The NACE limit of 22 HRC is the well known number, but it applies to sour service specifically. In general high strength fastening the practical thresholds are around 32 HRC for risk of hydrogen assisted cracking and 35 to 39 HRC for the ISO 898-1 property classes 10.9 and 12.9, which is exactly why those classes are effectively banned on many oil and gas specifications regardless of coating.
Plating is the most common self-inflicted cause. Electroplated zinc on a high strength bolt generates hydrogen at the cathode during deposition, and some of it diffuses into the steel. The mitigation is a bake at around 200 degrees C within a defined window, usually within an hour of plating and for several hours depending on the strength and the coating thickness, which allows the hydrogen to diffuse out. It is a process that must be qualified and audited, not ordered as a note on a drawing, which is why zinc flake coatings, applied by dip-spin and cured rather than electrodeposited, have become the default for high strength fasteners in aggressive atmospheres.
For hydrogen service specifically, the specification usually ends up as B7M or B8M at a hardness cap, with no electroplating, with a documented bake if any coating is applied, and with process control to ASTM F1940 and mechanical verification to ASTM F519. The bolt is not the only item on that list, but it is the one with the smallest cross section and the highest stress.
5. Thermal relaxation and the hot re-torque
A joint that is perfect cold can be loose hot, for two independent reasons. The first is differential expansion. A carbon steel bolt has a coefficient near 12.2 times ten to the minus six per kelvin and an austenitic stainless flange near 16.5, so the flange grows faster than the bolt as the joint heats. On a 150 millimetre grip and a 200 kelvin rise, the flange moves about 0.13 millimetres further than the bolt, which on the stiffness of a single M24 stud is worth tens of kilonewtons of preload, absorbed by compressing the gasket further or, if the gasket has no more travel, by over-stressing the bolt.
The second is creep in the gasket and in the bolt. Graphite and PTFE based gaskets continue to thin under load at temperature, spiral wound gaskets take a set, and the bolt itself relaxes by a few percent through stress relaxation. Published figures for preload loss over the first thermal cycle at 400 degrees C commonly run from 20 to 40 percent, and a joint that started at 60 percent of yield can end up below the seating load with nothing visibly wrong.
The practical answer is a hot re-torque: bring the joint to operating temperature, hold for a defined period, usually 24 hours, allow it to cool if the procedure requires, and re-tighten to a stated fraction, often 70 to 100 percent of the initial torque, following the same pattern. This is standard practice on steam and hot oil systems and it is routinely omitted on valve flanges because the flange is considered part of the valve rather than part of the piping. It should be in the commissioning procedure, with the torque values and the sequence written down, because a re-torque done by feel is a re-torque that introduces scatter.
The other option is to design the relaxation out: use a gasket with a low creep relaxation, use longer and therefore more elastic studs, use disc springs under the nuts to store more travel, or use live loading. All of them add cost, and all of them are cheaper than a shutdown.
6. Class 900 and above: ring joint, rigidity and twelve bolts
As pressure class rises the joint stiffens and the failure mode changes. Class 150 and 300 flanges are relatively flexible, the gasket does most of the work, and a joint can often be brought back by tightening. At Class 900 and above the flanges are thick, the bolts are large, and the gasket is often a ring joint: a metal octagonal or oval ring that seats on a tapered groove, metal to metal, with no gasket factor in the usual sense and no forgiveness at all.
A ring joint fails in two ways that a spiral wound joint does not. It fails if the groove is damaged, because there is no soft material to fill a nick, and it fails if the ring is reused after it has been compressed, because the contact has already yielded and the second seating is on a different geometry. The standard is ASME B16.20 and the ring should be replaced, not annealed and reused, except where a specific procedure allows it.
The bolt count matters too, and not in the direction people expect. Twelve bolts around a DN200 flange sound like a lot of redundancy until the arithmetic is done: the code load required about 1,825 square millimetres of B7, and twelve M27 studs provide about 5,500. The flange is therefore bolt rich, which means a properly made joint runs at 30 to 50 percent of yield, and it also means that a single missed bolt changes the load distribution around the gasket rather than simply removing one twelfth of the capacity. The gasket does not average; it leaks at the point of least contact stress.
Large diameter flanges have a further problem that is a matter of geometry rather than strength. Above about NPS 24, flange rotation under bolt load becomes significant, the gasket contact stress is higher at the outer edge than at the bore, and the joint can leak at the inside diameter while appearing tight outside. This is why large valves often use a ring joint, a confined gasket or a welded connection, and why the bolt tightening sequence matters more, not less, as the flange grows.
7. Seawater, stainless and galling
Two atmosphere driven failures deserve their own paragraph because both look like something else. The first is galvanic corrosion: a stainless bolt in a carbon steel flange, wetted by sea water or by spray, becomes the cathode and the flange becomes the anode, and the flange, which is the expensive part, corrodes preferentially around the bolt hole. The joint then leaks not because the bolt failed but because the flange face has been eaten away beneath it. Isolating washers, a coating system that covers the flange face and the bolt, or simply matching the materials, all address it.
The second is galling. Austenitic stainless nuts and studs, particularly in the coarse thread sizes used on large flanges, can cold weld to each other during tightening, and the result is a nut that seizes halfway down, a torque reading that means nothing, and a stud that has to be cut off. Anti-seize compound on the thread and under the nut is the standard prevention, and it also stabilises the friction coefficient, which is the same thing as saying it makes the torque mean something.
Reuse is the third trap. A bolt that has been tightened, heated and loosened has yielded threads, a changed surface and a changed friction coefficient, and its preload for a given torque is no longer the value in the table. Most operators permit one reuse for low temperature non-critical joints and forbid it for high temperature, hydrogen and sour service. Whatever the rule, it has to be written, because the decision is always taken in the field by somebody with a bucket of old studs.
8. The tightening procedure that actually works
The procedure is boring and it is where most joint failures are created. Clean the threads and the nut faces, apply the specified lubricant to the thread and to the face under the nut, and use the same product on every bolt. Bring all nuts to hand tight, then to a low value, and only then to the target, in a star or crisscross pattern that moves around the flange so that opposite bolts are always tightened together. Three passes at roughly 30, 70 and 100 percent of target is the usual pattern, and the final pass should go around the flange twice, because tightening one bolt unloads its neighbours.
Check the geometry before any of that. The stud must engage a full nut, with at least one thread protruding beyond the nut and preferably two or three; a stud that is flush with the nut is carrying its load on the last incomplete thread. The nut face must bear on the flange or on a hardened washer, not on a paint film. The gasket must be centred, and on a spiral wound gasket the centring ring must be on the outside of the bolt circle, not trapped under it. And the flange faces must be parallel: a joint pulled up with a 2 millimetre gap on one side has a bending load in every bolt before the pressure arrives.
Thread | Stress area, sq mm | Preload at 60 percent of B7 yield | Torque, K = 0.20, dry | Torque, K = 0.16, lubricated |
|---|---|---|---|---|
M16 | 157 | 68 kN | 218 N m | 174 N m |
M20 | 245 | 106 kN | 425 N m | 340 N m |
M24 | 353 | 153 kN | 735 N m | 588 N m |
M27 | 459 | 199 kN | 1075 N m | 860 N m |
M30 | 561 | 244 kN | 1461 N m | 1169 N m |
M36 | 817 | 355 kN | 2553 N m | 2042 N m |
Read that table as an illustration of the method rather than as a specification, and note how much the torque moves for a change of K from 0.20 to 0.16 at the same preload: a 20 percent difference on the wrench for an identical result in the joint. That factor alone is larger than most of the refinements people argue about, and it is the reason a project should specify one lubricant, one application method and one calibrated wrench rather than a torque number copied from a catalogue.
Field symptom | Mechanism | What it is telling you | Correct action |
|---|---|---|---|
Weep at one clock position only | Uneven preload, gasket contact stress low at that point | Pattern or sequence was wrong, or one bolt was missed | Loosen all, re-torque in three passes with a proper pattern |
Leak that appears at temperature, not cold | Differential expansion and gasket creep | No hot re-torque, or a gasket with high creep relaxation | Hot re-torque to procedure, consider disc springs or live loading |
Bolt broken days after tightening | Hydrogen assisted cracking, plated or over-hard fastener | Hardness above 32 HRC or an unbaked electroplated coating | Replace the set with a controlled hardness grade, no electroplating |
Nut seized halfway down the stud | Galling of stainless on stainless | No anti-seize, coarse thread, high speed wrench | Cut and replace, apply anti-seize, reduce wrench speed |
Flange face corroded around the bolt holes | Galvanic couple between stainless bolt and carbon flange | Material mismatch in a wet or marine atmosphere | Match materials or isolate, review the coating system |
Joint tight on the outside, leaking at the bore | Flange rotation on a large diameter, gasket stress concentrated outside | Bolt load too high for the flange rigidity, or the wrong gasket type | Reduce target preload, use a confined or ring joint gasket |
9. A bolting clause you can paste into a specification
Flange bolting shall be supplied as studs with two nuts, in the grade stated, with material certificates to EN 10204 3.1 traceable to the heat number. Carbon steel valves in general service shall use ASTM A193 B7 with A194 2H nuts; valves in sour service to NACE MR0175 shall use A193 B7M with A194 2HM nuts, hardness tested to 22 HRC maximum on 100 percent of the studs and nuts; valves for service below minus 29 degrees C shall use A320 L7 or L7M; valves above 450 degrees C shall use the grade stated in the purchaser's specification, typically A453 Grade 660. No fastener shall exceed 32 HRC without the purchaser's written approval, and electroplated zinc coatings shall not be used on fasteners above 32 HRC unless a post-plating bake is specified, qualified and recorded. Where a corrosion resistant coating is required, zinc flake or PTFE based coatings applied by dip-spin shall be used, with a stated dry film thickness and a coefficient of friction declared by the CoF test method. Studs shall be long enough to give full nut engagement with a minimum of two threads protruding, and hardened washers shall be supplied where the flange material is softer than the nut. The tightening procedure, the lubricant, the target torque or tension, and the pattern shall be stated in the installation procedure, and a hot re-torque shall be specified for joints operating above 200 degrees C. A preservation and segregation requirement shall apply so that B7 and B7M are not mixed in the same crate, since the two are visually identical and the difference is a hardness value.
That last sentence is not a formality. B7 and B7M studs look the same, weigh the same and are marked on the end, and a mixed crate on a sour service project has happened often enough that many operators now require colour coding and separate packaging. J-VALVES supplies the valve and can supply the bolting with it, matched to the body material, the gasket type and the site atmosphere, with the grade, the coating and the certificates called out separately so that the two grades never travel together. The body range runs from 1/2 inch to 56 inch, DN15 to DN1400, ASME Class 150 through 2500, in carbon steel, stainless, duplex and nickel alloy, from minus 196 to plus 650 degrees C, tested at 1.5 times the rating on the shell and 1.1 times on the seat to API 598 and API 6D. Every shipment carries 3.1 certificates for the pressure parts and, on request, for the fasteners as well, and a quotation with the bolting schedule resolved comes back in three working hours.
10. Frequently Asked Questions
Why does a flange leak when the bolts are tight?
Because tight is not a number. Preload is what holds a joint, and torque only estimates it: the torque coefficient K ranges from about 0.12 to 0.25 with coating, lubrication and surface condition, so the same wrench setting can deliver preloads that differ by 30 percent or more between bolts. Add a pattern that unloads neighbours, a gasket that has taken a set, or a hot joint with no re-torque, and the contact stress at one point on the gasket falls below what is needed to seal. The joint then leaks at that point while every bolt feels tight.
What is the difference between A193 B7 and B7M?
Same chemistry, different heat treatment and acceptance. B7 is quenched and tempered to a minimum yield of 105 ksi, about 724 MPa, and is not hardness limited tightly enough for sour service. B7M is deliberately under-strengthened to a minimum yield of 80 ksi with a maximum hardness of 22 HRC, and every stud and nut is hardness tested. That is what NACE MR0175 requires in wet H2S service, because above 22 HRC the steel is susceptible to sulfide stress cracking. The two grades are visually identical, which is why they should never be packed in the same crate without colour coding.
How much preload is lost when a flange joint gets hot?
Commonly 20 to 40 percent over the first thermal cycle at around 400 degrees C, from two causes. The flange and the bolt have different expansion coefficients, roughly 16.5 against 12.2 times ten to the minus six per kelvin for a stainless flange and a carbon bolt, so a 150 millimetre grip moves about 0.13 millimetres apart over a 200 kelvin rise. At the same time the gasket continues to creep and the bolt stress relaxes. A hot re-torque after 24 hours at operating temperature, back to 70 to 100 percent of the initial torque using the same pattern, is the standard remedy.
Can I use stainless bolts on a carbon steel valve flange?
Yes, but understand what you are creating. A stainless stud in a carbon steel flange in a wet or marine atmosphere makes the flange the anode of a galvanic couple, and the flange face corrodes around the bolt holes, so the joint leaks for a reason that has nothing to do with the bolt. If stainless bolting is required, coat the whole joint, use isolating washers, or upgrade the flange material. The other trap is grade: annealed B8 has a yield near 200 MPa, a third of B7, and will relax on a joint designed for a high strength stud. Ask for B8M Class 2 if stainless is needed at strength.
Why do bolts break after tightening rather than during?
That delay is the signature of hydrogen assisted cracking. Atomic hydrogen, from electroplating, from cathodic protection or from the process, diffuses into the steel, collects at inclusions and grain boundaries, and a crack initiates and grows under a load that is well below the tensile strength, sometimes within hours and sometimes within days. Hardness is the controlling variable, which is why limits of 22 HRC for sour service and around 32 HRC in general high strength fastening exist, and why electroplated zinc on a high strength bolt requires a qualified post-plating bake.
