Flange Joint Integrity — Torque, Tension and the Gasket Window

The site situation

An 8″ Class 300 raised-face joint on a hot hydrocarbon line has wept twice since start-up. Both times the fitter went out with a torque wrench and pulled the bolts up harder, and both times it stopped weeping for a few weeks. Today it is weeping again, and now one stud has snapped.

The supervisor's diagnosis is the one everybody's hands agree with:

"It's not tight enough. Put a cheater bar on it."

Guess first: of the bolt load you put in with a torque wrench, how much of the torque you apply actually ends up as bolt stretch? Half? A quarter?

About 13 %. Roughly 48 % of your effort is burned as friction under the nut face and 39 % in the threads. Which is why two identical-looking studs, torqued identically by the same person on the same joint, routinely end up ±30 % apart in actual tension — and why "tighter" is a control input you cannot see the output of.

And there is a second, bigger problem with the cheater bar: a gasket has a maximum as well as a minimum. There is a window, and it is possible to be above it.

The misconception, stated plainly, then dismantled

"Tighter is better. If it leaks, it wasn't tight enough."

This survives because the first thing you try often works. Most under-tight joints do respond to more torque, so the rule gets reinforced. It is a rule that trains you right up to the point where it destroys the joint.

Follow the load path. Nothing in a bolted joint is direct:

torque on the nut  →  (13 %)  →  bolt tension  →  gasket seating stress
        →  gasket conforms, fills the surface finish, seals
        →  internal pressure pushes the flanges apart (hydrostatic end force)
        →  what is LEFT over is the residual gasket contact stress
        →  the joint leaks when residual stress < the gasket's minimum

Every arrow in that chain loses something, and every loss is uncertain. So "tighter" is really three separate claims — that more torque gives more tension, that more tension gives more gasket stress, and that more gasket stress gives a better seal. The first two are roughly true with large error bars. The third one is false above a limit.

A gasket has two numbers, not one:

• Minimum seating stress (y or S_min) — below it, the gasket never conforms to the flange surface and the joint leaks from day one.
• Maximum allowable stress (crush, S_max) — above it, the gasket is destroyed: a spiral wound is compressed solid and loses all springback, a sheet gasket extrudes and crushes, a PTFE gasket cold-flows out of the joint over weeks, an RTJ ring flattens and damages its grooves.

Between them is the window. A joint works when the whole scatter band of your bolt load lands inside that window, at every point around the circumference, after the pressure has pushed back and after the joint has relaxed. Stated that way, "more is better" is obviously the wrong mental model — you are aiming at a target, not pushing in a direction.

Bolt tension from torque:            T = K · D · F          (so F = T / (K·D))
Gasket seating stress:               S_g = n · F / A_g
Hydrostatic end force:               F_H = (π/4) · G² · P
Residual gasket stress:              S_res = (n·F − F_H) / A_g

Indicative windows (teaching values — always use the gasket maker's data or ASME PCC-1 Appendix O):

Note the RTJ row. The same bolt load produces about three and a half times the stress on an RTJ as on a spiral wound (12.7 / 3.5 = 3.6), because the contact strip is 3.5 mm instead of 12.7 mm. Applying a spiral-wound torque table to a ring joint is one of the classic ways to wreck a flange.

Drive the whole chain yourself — pick the flange, the gasket and the torque, set the nut-factor uncertainty, and watch the tension band and the gasket stress band move against the window: ▶ open the interactive: construction flange integrity calc

Symbol key — every symbol on this sheet

• T — applied torque at the nut · N·m
• K — nut factor ("torque coefficient"): the empirical fudge that converts torque to tension · –
• D — nominal bolt diameter (not the thread pitch diameter) · mm
• F — tension in one bolt · N, kN
• n — number of bolts in the joint · –
• A_s — bolt tensile stress area (the thread root area that carries the load) · mm²
• σ_b — bolt stress = F / A_s; targets are quoted as a % of bolt yield · MPa
• SMYS_b — bolt yield; ASTM A193 B7 up to 2½″ is 725 MPa · MPa
• A_g — gasket contact area = π · G · w · mm²
• G — gasket mean (effective) diameter — also the diameter the end force acts on · mm
• w — gasket contact width · mm
• S_g — gasket seating stress = n·F / A_g · MPa
• S_min, S_max — minimum seating and maximum (crush) stress: the two edges of the window · MPa
• F_H — hydrostatic end force = (π/4)·G²·P — pressure trying to open the joint · N, kN
• P — internal pressure · barg, Pa
• M — external bending moment applied to the joint by the pipe · N·m, kN·m
• Z_g — gasket-annulus section modulus = π·R²·w, with R = G/2 · mm³
• ΔS_g — one-sided gasket stress change from M, = M / Z_g · MPa
• m, y — ASME VIII Appendix 2 gasket factors: maintenance factor and seating stress · –, MPa

Why torque is a poor proxy for tension

T = K · D · F        K ≈ 0.10–0.12 (PTFE)   0.12–0.15 (moly grease)
                     0.13–0.18 (nickel anti-seize)   0.20–0.35 (dry / as-received)

K is not a property of the bolt. It is a property of the bolt, the nut, the washer, the flange face under the washer, the lubricant, how evenly the lubricant was applied, whether the threads are new, whether it is raining, and the speed of the wrench. A dry stud and a moly-greased stud at the same torque differ by a factor of two in tension. Within a single well-lubricated, well-controlled joint, achieved preload still scatters about ±25–30 %.

T / F  =  p/(2π)              thread lead — the only useful term     = 0.449 mm
       +  μ_t · r_t · sec α   thread friction                        = 1.413 mm
       +  μ_n · r_n           nut-face friction                      = 1.716 mm
                                                          total      = 3.578 mm
K = 3.578 / 22.225 = 0.161

The alternatives, and what each buys:

Notice that all of them are worse than the arithmetic suggests, and that the good methods stop measuring torque and start measuring stretch. Tension is the physical quantity. Torque is a proxy with a friction-shaped error on it.

Sequence, relaxation and the loads that arrive later

Why crossed patterns exist

Tightening one bolt does not only load that bolt. It squashes the gasket locally, which lets the flange move, which unloads the bolts already tightened nearby — the phenomenon is called elastic interaction, and on a soft gasket the first bolts can lose 30–50 % of their load by the time the last bolt is done.

• Going round the circle in order, 1-2-3-4…, you tilt the flange progressively. Each new bolt closes the gap it can reach, and the gasket ends up wedge-shaped: highly compressed where you finished, barely seated where you started. This is the "chasing" joint that never settles.
• Going across the diameter — 1, opposite, 90° away, opposite — each bolt's local squash is balanced by the next one on the far side. The flanges come down parallel, and the stress distribution rises evenly.

ASME PCC-1 Appendix F sets out the pattern that industry has converged on:

Pass 1   hand-tight / snug, cross pattern      — check flange gap is even all round
Pass 2   30 % of target torque, cross pattern
Pass 3   60 % of target torque, cross pattern
Pass 4   100 % of target torque, cross pattern
Pass 5   100 %, ROTATIONAL (circular) pass — keep going round until no nut moves

Pass 5 is the one people skip and the one that does the most work: it is what removes the elastic interaction losses left over from passes 2–4. It is also free.

Watch the two patterns build up — the crossed sequence colouring the gasket evenly, versus the sequential one tilting the flange into a wedge, and then a moment prising one side open: ▶ open the interactive: construction flange integrity 3d

Relaxation, flange rotation and external loads

Three things attack the load after the fitter has gone home.

Short-term relaxation (embedment). Surface asperities on the bolt threads, nut faces and gasket bed down within minutes to hours. Expect 10–30 % loss, more with soft gaskets and coated bolts. This is why a re-torque after 4–24 hours, at ambient, is standard on critical joints.

Flange rotation. The bolt circle is outside the gasket, so bolt load applies a moment to the flange ring and it dishes — the gasket sees more stress at its outer edge and less at its inner edge. Weak (Class 150, large-bore) flanges rotate most, which is precisely where the gasket stress is already marginal. It is also why narrow gaskets close to the bolt circle behave better.

Creep and thermal relaxation. Graphite and PTFE creep; the bolts, flanges and gasket all expand at different rates and the flange is hotter than the studs during a ramp. Hot re-torquing recovers this — but it is a hazardous operation on a live line and is only done where the procedure explicitly permits it, at a controlled reduced pressure, never as an improvisation.

External pipe loads are the link back to piping stress. A bending moment M applied to the joint does not act on the gasket uniformly — it adds on one side and subtracts on the other:

Z_g = π · R² · w          (section modulus of the gasket annulus, R = G/2)
ΔS_g = M / Z_g            → one side S_g + ΔS_g, the other side S_g − ΔS_g

This is the quiet killer. Your stress analysis said the flange was fine because the nozzle was within allowable. But the joint does not care about the nozzle allowable — it cares whether the unloaded side is still above minimum seating. A moment that is perfectly acceptable to the pipe can still take one side of the gasket below its seating stress, and a joint that leaks on one side leaks.

Worked example — the 8″ Class 300 joint that keeps weeping

8″ Class 300 RF, spiral wound with graphite filler, 12 × 7/8″ A193-B7, moly-greased (K = 0.16), 40 barg operating, target 50 % of bolt yield.

Step 1 — target tension and torque

A_s(7/8″–9 UNC) = 298 mm²      σ_b = 0.50 × 725 = 362 MPa
F  = 362.5 × 298 = 108.0 kN per bolt
T  = K·D·F = 0.16 × 0.022225 × 108,000 = 384 N·m

Step 2 — nominal gasket stress

G = 247.2 mm,  w = 12.7 mm  →  A_g = π × 247.2 × 12.7 = 9,863 mm²
S_g = 12 × 108,000 / 9,863 = 131 MPa        window 70 – 240 → comfortable

Step 3 — now apply the real world, one layer at a time.

±30 % torque scatter          bolt total 907 – 1,685 kN
relaxation                    ×0.80 on the BOLT LOAD (embedment relaxes the studs, not the
                              hydrostatic end force — take it off first, not off the answer)
hydrostatic end force         F_H = π/4 × 0.2472² × 40e5 = 192 kN   (always subtracts)
external moment 15 kN·m       Z_g = π × 123.6² × 12.7 = 6.10e5 mm³ → ΔS_g = ±24.6 MPa

(Order matters here and it is easy to get wrong. Relaxing the net gasket stress by 20 % instead of the bolt load flatters every number by 0.2 × F_H/A_g ≈ 4 MPa — always in the unsafe direction.)

There it is. The nominal joint that looked comfortable at 131 MPa is at 61 MPa on its low side — below the 70 MPa minimum seating — and the unlucky-bolt case is at 30 MPa, well under half of what it needs. The joint weeps, on one side, intermittently, exactly as observed.

And now test the cheater bar. Suppose the fitter adds 40 % torque. Nominal bolt stress goes to 507 MPa — 70 % of yield, still legal for B7 — and the low side recovers to 103 MPa. It stops weeping, which "proves" the theory. But the lucky bolts are now at 91 % of yield, one of them is the one that snapped, and if this joint had been fitted with the RTJ the line was originally specified for, the 3.5 mm contact strip would have seen 476 MPa at the original torque, so 476 × 1.4 = 666 MPa on the ring, past its 620 MPa limit — permanent groove damage, and a joint that can never be made to seal again.

The correct fixes are the unglamorous ones: lubricate and re-use the PCC-1 pattern with a controlled K, add the rotational pass, re-torque after bedding-in, and go back to the stress model and reduce the 15 kN·m — because a quarter of the problem is not in the joint at all.

Common pitfalls and the leak-cause ranking

Typical leak causes, ranked

Roughly the order that joint-integrity surveys keep finding:

1. Uneven or insufficient bolt load — no pattern, no passes, no rotational pass, dry bolts.
2. Flange face damage or wrong surface finish — radial scores across the serrations, rust, paint on the face, the 3.2–6.3 µm Ra spiral serrated finish polished off.
3. Wrong, damaged or misaligned gasket — wrong class, wrong bore, two gaskets, a gasket installed off-centre so its inner ring sits in the flow.
4. Flange misalignment forced closed with the bolts — parallelism and offset outside the PCC-1 limits, so a permanent moment is locked into the joint before it ever sees pressure.
5. External loads and thermal cycling — the moment case above; also every start-up and shutdown ratcheting a little more load out of the joint.
6. Over-compression — crushed sheet gaskets, cold-flowed PTFE, spiral wounds compressed solid.
7. Bolting hardware — reused or galled studs, no hardened washers, inconsistent lubricant, wrong grade.

Note that (1), (4) and (6) are all load distribution problems, not load magnitude problems. That is the whole point.

The pitfalls themselves

• Using one torque table for every gasket type. The gasket, not the bolt, sets the target.
• Torquing to a value from a chart without knowing what K it assumed. A dry-bolt chart used on lubricated studs over-tensions by up to 2 ×.
• Skipping the final rotational pass, then blaming the gasket.
• Pulling a misaligned flange together with the studs — the pipe strain does not disappear, it becomes a permanent moment sitting inside your gasket stress budget.
• Re-using studs and nuts "because they look fine". Thread condition is the calibration.
• Checking bolt stress against yield and stopping there — the gasket crush limit usually bites first, and nothing warns you when it does.
• Forgetting that pressure subtracts: a joint that seats beautifully at 0 barg can be below minimum seating at operating pressure.
• Treating "the nozzle load passed the code check" as "the flange joint is fine". Different check, different failure mode.

Outcome

• The load path is torque → tension → gasket seating stress → residual stress after the hydrostatic end force. Only about 13 % of torque becomes stretch; the rest is friction.
• Every gasket has a window, S_min to S_max. Design and assembly must put the entire scatter band inside it, all round the joint, after pressure and after relaxation.
• Torque control scatters preload ±25–35 %; turn-of-nut ≈ ±15 %, tensioners ≈ ±10 %, measured elongation ≈ ±5–10 %. Anything that changes friction changes tension.
• Crossed-pattern tightening in 30/60/100 % passes plus a final rotational pass (PCC-1 Appendix F) is what removes elastic interaction; sequential tightening builds a wedge.
• External moments split the gasket stress: ΔS_g = M / (π·R²·w), one side up, one side down. The low side is what leaks — and that is a piping-stress result, not a construction result.
• Interactive: ▶ open the interactive: construction flange integrity calc — tension band vs the gasket window, with a moment slider. 3D: ▶ open the interactive: construction flange integrity 3d — pattern and moment shown on the gasket itself.

Open items

• Add ASME PCC-1 Appendix O target-stress tables and the joint-stiffness (turn-of-nut) route.
• Extend the explorer with bolt-stress % of yield limits per material (B7, B7M, B8M, L7) and a temperature derate.
• Worked flange-rotation calculation (Appendix 2 / Taylor-Forge) to show the inner-edge unloading explicitly rather than as a note.

Educational teaching material — simplified and illustrative. Not for engineering design use.
© EPiC Circle · provided by PipeAxis