How to Bolt Up a Flange So It Seals the First Time

Why this matters

A flange that leaks on start-up costs you the start-up, and on a customer site it costs you the reputation you built getting the rest of the job right. The reflex when it weeps is to go around it again with the wrench, which works often enough to become a habit and hides the reason it leaked.

Most first-time flange leaks are decided before a single bolt is tightened. The two faces were not parallel, or they were offset, or the bolt holes did not line up and somebody pulled them together with the studs. The tightening sequence then distributes whatever load it can onto a gasket that is being asked to seal a wedge-shaped gap. This procedure puts the fit-up check first and treats the bolts as something you use only after the joint already fits.

1. Make the joint dead, prove it, and hold it up

Do this before anything else, and do not compress the steps.

• Isolate the line and lock out the mechanical and stored energy sources under 29 CFR 1910.147. Prove zero pressure by cracking a vent or drain at the joint with the isolation locked, standing to the side of the outlet, not by reading a gauge somewhere upstream.
• If there is any conductor, heat trace, instrument lead, or motor termination on the equipment, that part is electrical work: de-energize and verify absence of voltage under 29 CFR 1910.333(b)(2), and prove the tester live-dead-live per NFPA 70E-2021, 120.5, which binds you through your employer's electrical safety program or your contract, not on its own.
• If the line was hot, let it cool before loosening. Back the bolts off in the crossing pattern with your body out of the split plane, and rig or crib the pipe and any valve or component hanging off the joint before the last bolts come out. The flange is carrying that weight until it is not.
• If the joint is on a metallic water or process line that may be serving as a bonding path, install a bonding jumper across the opening before you separate it, per the requirements of NEC Article 250 in the edition your authority having jurisdiction has adopted. Separating that line can put fault current across your hands.

Old gasket residue is an inhalation hazard until proven otherwise. Sheet gasket material on older industrial equipment may be asbestos-containing. Do not power wire-brush, grind, or dry-scrape unknown gasket residue. Treat it as presumed asbestos-containing material until it is assessed, use wet methods and hand scraping, and follow 29 CFR 1910.1001 in general industry or 29 CFR 1926.1101 in construction, with respiratory protection under a written program per 29 CFR 1910.134. A glove does nothing about this route.

2. Read both faces before you look at anything else

Clean both flange faces down to bare metal in the seating band and look at them in good light.

What matters is a defect that crosses the seating band radially. A circumferential scratch that follows the serrations is inside a groove the gasket is already going to fill. A radial scratch is a channel from the bore to the atmosphere, and no amount of load closes it. So is a nick on the raised-face edge, and so is a corrosion pit deep enough to catch a fingernail.

A face with a defect that crosses the band gets re-faced or replaced. That is a machining decision with an acceptance standard behind it, and the standard belongs to the piping code or the equipment manufacturer, not to the tech judging it by eye.

3. Check fit-up, and treat it as a gate

This is the step that gets skipped, and it takes about three minutes. With no gasket in place, bring the flanges together loose and measure four things.

Face separation at four points, at 12, 3, 6, and 9 o'clock, with the same feeler technique at each. This is the parallelism check.

Centerline offset, by laying a straightedge across the two outside diameters at two positions ninety degrees apart. Offset shows up as a step.

Rotational alignment, by whether every bolt drops through both holes by hand. If you need a drift pin to line the holes up, use one and keep your fingers clear of the hole while it is in there; never align holes with a finger.

Gap versus gasket thickness. The joint should close on the gasket, not on a gap you intend to pull out with the studs.

The tolerance on all four belongs to the adopted piping code or the project specification, and where a rotating machine is on one end of the joint, to the machine manufacturer's alignment limits. What is universal is the rule: if the joint does not fit within its stated tolerance, fix the pipe run, not the bolt torque. A joint pulled into position with studs stores that bending load permanently, and it comes back out of the gasket's seating budget on the far side of the flange.

4. Fit the gasket dry and concentric

Right material for the service, right size for the face, undamaged. It sits on the raised face without overhanging into the bolt circle, because overhang is contact area that dilutes your seating stress without sealing anything.

Do not use a gasket compound or spray to hold a gasket in place unless the gasket manufacturer specifies one for that material. On many materials it changes the friction between gasket and flange and lets the gasket move as the joint is loaded.

5. Lubricate the threads and nut faces, where the spec allows it

A large share of the torque you apply is spent on friction under the nut face and in the threads, and dry, rusty, or inconsistently lubricated fasteners scatter the load you actually get from a given torque. A consistent lubricant on threads and nut faces is what makes the torque-to-load relationship repeatable across eight bolts.

Two carve-outs, both hard. Where the specification forbids lubricant, or the service is oxygen or another strong oxidizer, do not lubricate: hydrocarbon-based products in an oxygen system are an ignition source, and that call routes to the system owner and the lubricant manufacturer's compatibility data, not to the truck.

6. Load in passes, in the crossing pattern

Snug all bolts hand-tight plus a small consistent amount in the crossing pattern first, so the gasket is held flat before any bolt takes real load. Then work up in passes rather than taking any bolt to final value on its own, and finish with a pass around the circumference in order. A sibling article owns the pattern and pass structure in detail; the reason it belongs in this procedure is that every pass is redistributing load across a joint whose geometry step 3 already made correct. On a joint that did not pass step 3, the pattern is redistributing a problem.

7. Measure the compressed gasket, and accept or reject on that

This is the acceptance check, and it is the only one that reads what the gasket experienced.

Worked example: the joint that was going to leak before anyone tightened it

A 6 in flange pair on a horizontal run, 8 bolts, bolt circle measured at 9.5 in, gasket 0.125 in thick. Illustrative values.

Face separation, loose, no gasket, at four points:

Three readings inside 0.003 in of each other and one 0.041 in wider than the tightest. Across a 9.5 in bolt circle that is an angular mismatch of about 0.0043 radians, roughly 0.25 degrees. It looks like nothing. It is not nothing: to close it, the bolts at 6 o'clock have to bend the pipe run 0.041 in, and every pound of load spent bending the run is a pound not clamping the gasket at 12 o'clock.

The fitter's instinct is to run the 6 o'clock bolts down first and let the rest follow. Suppose the crew does exactly that and gets the joint together. Compressed gasket thickness afterward, four points:

Read the 6 o'clock row. That quarter of the gasket has taken about a fifth of the compression the tightest quarter has. Whatever seating stress the material needs, the joint is nowhere near uniform, and the low-stress quarter is where it will weep on the first pressure cycle. Note also the direction of the error relative to the fit-up: the wide gap was at 6 o'clock and the light compression ended up at 6 o'clock too, which is the confirmation that the fit-up defect, not the tightening, produced the result.

The correct move is to back the joint off, find why the run sits low at that end (a hanger set at the wrong height, a spool made short, a support that settled), correct it, and re-check separation. On this joint it was a hanger rod that had been let out to clear a conduit. Ten minutes on the hanger; the flange then measured within 0.004 in across all four points and compressed within 3 percentage points all around.

The failure mode if you skip it: the joint holds hydrostatic test at ambient, because a cold static test is the easiest thing you will ever ask of it, and then weeps at 6 o'clock on the first hot cycle when relaxation takes a fraction of the load away from a face that never had margin there. The callback reads as a bad gasket.

How to verify you got this right

• Compressed thickness spread across the four points inside the gasket manufacturer's stated allowance. If they do not publish one, uniformity is still the signal: a spread as wide as the example above is a reject on its own.
• Every nut drawn down with no stud bent and no drift pin left carrying load. A drift pin is an alignment tool, not a bolt.
• Paint-mark nut to flange after final pass. It costs seconds and it turns the next visit's question, did it relax or did it rotate, into a look rather than a debate.
• Record grip length, gasket material, thickness, and the four compressed readings on the job. A joint record with only a torque number on it cannot tell the next tech anything about why it held.

References

• 29 CFR 1910.147, control of hazardous energy, for isolation and stored energy before opening a bolted joint
• 29 CFR 1910.333(b)(2) for electrical isolation, with NFPA 70E-2021, 120.5 for the live-dead-live proving sequence where your employer's electrical safety program adopts it
• 29 CFR 1910.1001 (general industry) and 29 CFR 1926.1101 (construction) for asbestos, with respiratory protection under a written program per 29 CFR 1910.134, where old gasket residue is presumed asbestos-containing
• NEC Article 250, in the edition your authority having jurisdiction has adopted, for bonding continuity across a separated metallic piping section
• ASME PCC-1, in the edition adopted by your project specification, for flange joint assembly including fit-up tolerance and pass structure; it binds through that specification rather than on its own
• See related: The Torque Sequence That Matters; How a Gasket Actually Seals; Why a Gasket Fails