How to seal milled aluminum enclosure corners

Seal a bolted milled-plate box with one continuous O-ring line and a dab of sealant at each three-plate corner. Cut a face groove around the front and back openings that turns 90° into end-face grooves on the side plates, so all four corner joints share a single gasket path. For 0.070 in [1.78 mm] cord, a groove 0.052 in [1.32 mm] deep and 0.106 in [2.69 mm] wide gives 26 % squeeze at 70 % fill. Each corner where three plates meet gets a 4 mm [0.16 in] dab of RTV worked into the groove crossover. That clears IP56 with margin: the jet test delivers about 0.9 bar [13 psi] of stagnation pressure, and a compressed face seal holds tens of bar from either direction.

What IP56 actually asks of the seal

IEC 60529 tests the second digit 6 with a 12.5 mm nozzle at 100 L/min, held 2.5 to 3 m [8.2 to 9.8 ft] from the box. The spray runs at least 3 min, from every direction worth aiming. Run the arithmetic: 100 L/min through a 12.5 mm bore leaves the nozzle at 13.6 m/s [45 ft/s]. Where that jet stagnates on a flat plate it develops about 0.9 bar [13 psi]; a seam at an angle sees less.

The first digit is the easier one. Digit 5 is dust-protected — dust may enter if it cannot interfere with the internals — and only digit 6 is dust-tight. Nothing in IP56 involves immersion: that is digit 7, 1 m [3.3 ft] for 30 min. The full digit meanings are in the IP rating chart.

One line on markings: an IP code on a finished box is a manufacturer rating unless a lab has listed it, and ours are rated by the manufacturer, not UL listed. So plan your own leak check; the 5 psi soap test at the end of this article is that check.

One gasket line instead of six

The trap in a six-plate box is sealing every joint separately: six gaskets, eight three-plate intersections, and every one of them a leak you have to hunt. Give the box two openings instead and seal each with one loop. On a box whose front and back plates bolt to a heatsink chassis, that is exactly what you have. Two openings, four corner seams each, all carried by one ring.

Machine it so the ring never crosses a butt seam on its own. Mill the groove into the mating face of the front and back plates, right around the perimeter, and mill matching grooves into the end faces of the side plates. At each corner the ring leaves a face groove, crosses the seam, and enters an end-face groove. Miter the groove ends at 45° and run each one 0.015 in [0.38 mm] past the theoretical intersection, so rubber fills the crossover instead of air.

Register the plates while you are at it. Dowels or a tongue and groove edge keep the grooves aligned within about 0.010 in [0.25 mm]. Beyond that, the ring walks out of the corner as the bolts come down.

If the design pushes outside air through a heatsink, treat that path as a tube through the box. Give the duct its own ring where it passes each wall — same cord, same groove numbers — and the airflow never touches the electronics volume. Draw the sealed boundary on the layout: the IP claim covers that volume and ends at the duct wall. If the intake opening itself must carry a code, the weakest part sets the assembly's rating.

Groove numbers that hold

These figures come from Parker's O-Ring Handbook (ORD 5700) for a static face seal in 70 Shore A rubber; a named compound's datasheet overrides them.

Two checks, shown: squeeze = (0.070 − 0.052) ÷ 0.070 = 26 %, inside the band. Fill = (π/4 × 0.070²) ÷ (0.106 × 0.052) = 0.00385 ÷ 0.00551 = 70 %, under the 85 % ceiling.

Compound by duty: EPDM for water and washdown, -50 to 150 C; silicone for wide temperature swings, -60 to 200 C; nitrile where oil is present, -30 to 120 C.

Worked example, a 12 × 8 × 4 in [300 × 200 × 100 mm] body in 0.375 in [9.5 mm] 6061 plate — plate thickness has its own guide in how thick a milled plate should be. Each opening carries about 39 in [1 m] of groove. The ring pushes back at roughly 25 lbf per inch of path, so one loop resists with 39 × 25 ≈ 975 lbf [4.3 kN]. Twenty-four M5 cap screws at 40 mm [1.6 in] pitch, torqued to 20 in-lb [2.3 Nm], develop about 500 lbf each. That is 12,000 lbf total, twelve times what the ring needs, with margin to keep the flange flat between screws; flatness control between the screws is worth a read before you fix the pitch.

Building the box, step by step

1. Lay out one ring loop per opening in CAD, groove centerline at least 0.10 in [2.5 mm] from any plate edge. Miter the groove ends 45° at each corner and extend them 0.015 in [0.38 mm] past the theoretical intersection.
2. Mill the grooves to ±0.002 in [0.05 mm] on depth, both groove families in one setup per plate where the machine allows. Break the edges at 0.010 in [0.25 mm] and finish the lands at 32 µin Ra [0.8 µm] or better.
3. Anodize before assembly. Type II grows a surface about 0.0003 in [8 µm], moving squeeze by under one point. Hardcoat builds 0.002 in [0.05 mm] and adds about three points of squeeze, so cut the groove 0.002 in deeper first, or machine the groove after coating.
4. Set the ring. A molded loop to your path is cleanest; otherwise join 0.070 in cord with cyanoacrylate and put the splice mid-run, at least 1 in [25 mm] from any corner. Light tension keeps the cord seated while you close the box.
5. Dab the corners. Work a 4 mm [0.16 in] RTV bead into each three-plate crossover with a plastic pick. RTV skins in 10-15 min at room temperature, so close the box inside that window; a 4 mm bead cures through in about 48 h.
6. Bolt the box with M5 cap screws at 40 mm [1.6 in] maximum pitch, blind-tapped so no fastener crosses the sealed wall. Torque 20 in-lb [2.3 Nm] in a cross pattern, two passes: 10 in-lb first, then 20.
7. After the cure window, run the soap test below before the box leaves your bench.

Prove it with 5 psi of air

Pressurize the sealed volume to 5 psi [0.35 bar] through a temporary fitting, brush soapy water over every seam and corner, and hold five minutes with zero bubbles. The test finds assembly errors — a pinched ring, a dry corner, a lifted splice — and 5 psi finds them loudly. The capability against a 0.9 bar jet is already in the squeeze numbers; this check proves you assembled it.

Use a regulator and a gauge, and never connect unregulated shop air. At 5 psi a 12 × 8 in plate carries about 480 lbf [2.1 kN]; at shop pressure it bows permanently and the grooves never line up again.

Mistakes that cost the rating

• Groove cut deep: at 0.058 in [1.47 mm] the squeeze falls to 17 %, below the band, and the first jet finds it. Hold depth to ±0.002 in [0.05 mm].
• Groove cut narrow: at 0.080 in [2.03 mm] wide the fill reaches 92 % and the plates stand off on displaced rubber.
• Splice near a corner: a glued joint within 1 in [25 mm] of a crossover sits where the ring must bend, and it lifts.
• Fasteners through the wall: every through-bolt is a leak path you then have to gland.
• Testing during cure: at 24 h the 4 mm dab is cured only halfway through, and it bubbles at the corner.
• Sharp groove edges: a burr nicks the ring on assembly; the 0.010 in [0.25 mm] break prevents it.

When to use another route

A flat rubber gasket is the low-tooling cousin of the groove: hand-cut to shape, trimmed in place, and good for IP56 on flat plates.

Decide by use. Panels that must open at IP56 get the continuous groove and the corner dabs. A box that closes for good gets welded seams and one machined lid face — sealing a large thin-walled lid to IP68 covers that geometry. And when milling six plates stops making sense, custom enclosures built from one piece arrive with your openings already cut, delivered in 10 business days counted from your approval of the drawing.

Enclosures for this job

Catalog sizes that suit the job this article answers, picked from its subject; open one for its full specifications.

Questions