What a Flanged Joint Needs in Order to Seal

Why this matters

A flanged joint does not seal because the bolts are tight. It seals because the gasket is held at a stress high enough to conform to both faces and stay there while pressure tries to push them apart. Stress is load spread over area, so two things that have nothing to do with the wrench decide whether you can get there: how much of the load reaches the gasket instead of bending something, and how much gasket area that load has to cover. A flange pair that is not flat and parallel puts a ceiling on gasket stress, and no torque value and no gasket upgrade gets you above it. That is why a joint can be assembled by a good tech, to the right value, in the right pattern, and weep anyway, and why the second attempt with a thicker gasket usually makes it worse.

Before breaking a flanged joint: isolate, drain, and confirm zero on a gauge at the joint rather than upstream. 29 CFR 1910.147 requires stored energy relieved, disconnected or restrained before service. Support the valve, spool or equipment the flange is holding before any bolt comes out, break the bolts on the far side of the joint first so the joint opens away from you, and stand out of the plane of the gap with face and body protection on. Treat any gasket in an older system as presumed asbestos-containing until it is proven otherwise. Do not dry scrape, wire brush or power-tool an old gasket off a face: use wet methods and hand tools under the requirements of 29 CFR 1910.1001 in general industry or 29 CFR 1926.1101 in construction, with the assessment made by the competent person those standards require. The route is inhalation, so gloves and safety glasses are not the control here.

What gasket stress is, and what this article assumes

Gasket stress is the total bolt load reaching the gasket, divided by the gasket's contact area. Both halves of that fraction move.

This article assumes you already got the bolt load right. Getting the bolt load right is genuinely hard, because torque is a proxy for it and a loose one, and a sibling article in this library owns that problem in full. Set it aside deliberately: even with exactly the correct clamp load in every bolt, a flange pair can be incapable of sealing. That is the subject here.

A gasket needs two different stress levels, and confusing them is common:

• A seating stress high enough to make the gasket flow into the face's surface texture on assembly. Below it, the gasket never conforms and the joint leaks from the first fill.
• A residual stress that remains after pressure applies its separating force to the joint. Internal pressure acting across the area inside the gasket pushes the flanges apart, and that load subtracts from what is holding the gasket. Below the residual the joint needs, it seals cold and lets go under pressure.

Those two properties are tabulated per gasket type as design factors in ASME Boiler and Pressure Vessel Code Section VIII, which reaches you through the vessel or piping code your authority having jurisdiction has adopted, and in the gasket manufacturer's own data for the specific material. The manufacturer owns the number for the gasket in your hand.

The gate: check flatness and parallelism before the gasket goes in

One check, done before assembly, with a straightedge and a feeler gauge:

Lay a straightedge across each face and try to pass a feeler under it, then bring the two flanges together dry, without the gasket or the bolts, and measure the gap at four points around the bolt circle.

That is it. If a feeler enters under the straightedge, the face is not flat, and no gasket will fix a face that is dished, warped, or has a radial score across the seating band. If the four gaps differ, the faces are not parallel, and the bolts will have to bend something to close the difference. Bolt load spent bending is bolt load that never reaches the gasket, and what does reach it arrives unevenly.

The gate is worth running because it takes a couple of minutes and it is the only check available while the fix is still cheap. After the gasket is in and the bolts are on, you cannot tell a misaligned pair from a well-aligned one without taking it apart again.

Outcome one: the pair that passed

Two pump flange pairs in the same plant, same size, same class, same gasket from the same lot at a nominal 1/16 inch thickness, same torque value, same crossing sequence, same wrench, same crew, same day. The only variable was the geometry the piping presented.

The first pair: no feeler entered under the straightedge on either face. Dry gaps at four points around the bolt circle sat within 0.005 inch of each other, effectively parallel.

After bolt-up, the flange gap measured at the same four points ranged from 0.054 to 0.056 inch against a nominal gasket thickness of 0.0625 inch. That is a compression of between about 10 and 14 percent, evenly around the circle. The gasket manufacturer publishes the compression range that material wants; this one landed inside it, all the way around.

The joint sealed on the hydrostatic test and stayed sealed through the first thermal cycle. Bolt load fell slightly on that first cycle, which is normal short-term relaxation and gasket creep, and the joint remained above the residual stress the service needed because it had started with margin evenly distributed.

Outcome two: the pair that did not

The second pair had the same everything except the pipe. The mating flange came up tilted: with the two faces pushed together dry, they touched at the bottom of the bolt circle and stood 0.080 inch apart at the top.

Nobody stopped. The bolts closed it, which is what bolts do.

After bolt-up, the flange gaps measured 0.052 inch at the bottom and 0.061 inch at the top. Against the same 0.0625 inch nominal gasket:

• Bottom: compressed 0.0105 inch, which is 16.8 percent.
• Top: compressed 0.0015 inch, which is 2.4 percent.

The bottom of that gasket is compressed about seven times as much as the top, from one bolt-up, at one torque value, by one person. Nothing about the wrench work was wrong.

Note also what the bolts spent themselves on. The dry non-parallelism was 0.080 inch and only 0.009 inch of it survived as a compression difference, so 0.071 inch of misalignment was taken out by bending the flanges and the pipe. That deformation is elastic, it is stored, and it is pulling on the joint for as long as the joint exists.

It leaked at the top, which is where 2.4 percent compression sits below any realistic seating requirement for a gasket of that class. That is the end of the range that fails first and fails visibly.

The other end of the range is the one that fails later. At 16.8 percent the bottom of the gasket is being taken toward or past its crush limit, where it loses the elastic recovery it needs to follow the faces when the joint heats, cools, and relaxes. A crushed section does not weep on the day; it stops tracking movement, and the joint that was fixed by a re-torque comes back a season later leaking somewhere new.

Why adding torque cannot rescue it. More torque at the top does not immediately raise stress at the top, because the first thing it does is bend the assembly further. Meanwhile the same additional load at the bottom lands on gasket already at 16.8 percent. You reach the crush limit at the bottom before you reach the seating stress at the top. The correction is to re-support and re-align the pipe so the faces meet parallel, then assemble with a new gasket. A gasket that has been through one bad bolt-up is not a candidate for the second attempt.

What would have changed the reading. If the dry gaps had been equal but a feeler had passed under a straightedge on one face, the problem would be flatness rather than parallelism, and the fix is machining or replacing that flange rather than moving the pipe. Same symptom at the joint, completely different work order, and the straightedge is what separates them.

What changes gasket stress and is not torque

Gasket contact area. Same bolt load over a wider gasket is a lower stress. This is the trap in "we will try a better gasket": swapping a narrow ring for a wider one at the same torque lowers the stress everywhere, and a joint that was marginal becomes a joint that leaks. Narrowing the contact area raises stress with no change to the bolts at all, which is why some gasket styles seal at loads a wider one cannot reach.

Full-face versus ring, on a flat-face flange. A flat-face flange, and cast iron in particular, needs a full-face gasket. A ring gasket on a flat-face pair puts the reaction inside the bolt circle and turns the bolts into a lever that bends the flange outboard of the gasket. On cast iron that cracks the flange, which is a failure with the line's contents behind it rather than a weep. Never chase a leak on a cast iron flange with more torque.

Face finish. The concentric or spiral finish specified for a flange face is a specification, not a manufacturing leftover. Too smooth and a soft gasket has nothing to key into; too rough and the gasket cannot fill the profile at its seating stress. The dimensional and finish requirements come from ASME B16.5 for the sizes it covers, invoked through the piping code your authority having jurisdiction has adopted.

Relaxation over the first hours and the first cycle. Bolted joints lose some load quickly through embedment and gasket creep. Where the gasket manufacturer specifies a re-torque, do it with the joint depressurised and cooled. Re-torquing a hot, pressurised joint is a controlled specialist procedure, not a field habit, and it is not what a service tech should be doing on a weeping flange.

References

• 29 CFR 1910.147, control of hazardous energy, for isolating, draining and supporting before breaking a flange; 29 CFR 1910.1001 (general industry) and 29 CFR 1926.1101 (construction) for presumed asbestos-containing gasket material, including wet methods and the competent person requirement
• ASME Boiler and Pressure Vessel Code Section VIII, for the gasket seating and residual design factors, and ASME B16.5, pipe flanges and flanged fittings, for dimensions and face finish, each in the edition invoked by the code your authority having jurisdiction has adopted
• Gasket manufacturer documentation for the specific material's seating stress, compression range, temperature and fluid limits, and re-torque instruction
• See related: What Torque Actually Controls and What It Does Not; The Torque Sequence That Matters; How to Set a Gasket So It Seals More Than Once