Sensor Housing Machining: Wall Thickness, Sealing Faces and Thread Coaxiality

A 12 mm inductive proximity sensor is a coil, an oscillator and a potted cable inside a metal tube. The tube is threaded at the front, sealed at the shoulder and bonded at the back. None of that is difficult geometry. All of it fails at the interfaces.

Sensor housing machining is judged in three places. Whether the potting stays bonded through a -40 to +85 °C cycle. Whether the O-ring still seals after the thread has been torqued into a bracket. Whether the sensing face stayed perpendicular to the thread axis after plating. A shop quoting from thread callout and bore alone will hit both diameters and still ship leaking bodies.

The three sealing interfaces

Almost every sensor body carries three seals, and only one is normally machined.

The panel seal is the one your drawing shows: a radial O-ring in a groove behind the thread, or a face gasket on the hex shoulder. An M12x1 body with a 2 mm cord radial seal needs a groove 2.4 mm wide and 1.5 mm deep, placed 0.2 mm behind the last thread flank so the ring is not cut as the body is screwed in.

The cap seal at the sensing end is the sensor maker's design — epoxy, a welded PPS cap, an overmolded front — but the bore that receives it is yours. A cap pressed into a bore 0.02 mm oversize will leak at 3 bar even with a full bead of adhesive, because adhesive cannot bridge a gap that varies around the circumference.

The connector seal is a molded M8 or M12 connector, an overmolded pigtail or a potted gland. For a standard M8 panel connector the housing needs a Ø8.1/8.2 mm hole, an anti-rotation flat at a defined angle, and a seat flat within 0.05 mm for the connector's own gasket.

Wall thickness and pressure

Potting compound shrinks 0.5 to 1.5 percent by volume as it cures. On a 1.0 mm brass wall over 20 mm the hoop stress is negligible, but the wall bows 0.02 to 0.04 mm — enough to spoil a thread fit if the thread is cut before potting.

Pressure rarely governs. Hoop stress is σ = p·d/(2t). At 10 bar (1 MPa) on a 16 mm bore with a 1.5 mm wall, σ = 5.3 MPa, nothing against brass at 200 MPa yield. At 100 bar the same wall sees 53 MPa: safe in brass, unsafe in POM, whose yield is near 60 MPa and which creeps under sustained load.

What sets the minimum wall is the thread. An M12x1 external thread has a minor diameter near 10.9 mm. With an 8.0 mm bore to clear the coil, the wall at the thread root is 1.45 mm. Workable in brass and 303 stainless, marginal in POM, and the reason almost no 10-bar sensor body is molded plastic.

See also: Battery Housing Machining: Gasket Grooves, Flatness and Leak Test Prep and Harmonic Drive Housing Machining: Keeping Three Bores on One Axis.

The groove that gets rejected

The radial (rod) groove sits in a cylindrical surface and seals against a bore. The face groove sits in a flat shoulder and seals against a flange. Radial grooves tolerate axial stack-up error; face grooves are easier to cut and inspect on a turned body.

Depth tolerance decides whether the seal works. For a 2 mm cord at 25 percent squeeze the nominal depth is 1.5 mm. Hold ±0.05 mm and squeeze stays between 22 and 27 percent, which every standard ring tolerates. Hold ±0.15 mm and a 70 Shore A ring at the shallow end will not compress enough at low pressure, while at the deep end it never touches both walls.

Three other numbers matter. Groove width 2.3 to 2.5 mm for a 2 mm cord, so the ring rolls into place but cannot extrude. Floor finish Ra 0.8 to 1.6 µm — not finer, because below Ra 0.4 µm a dynamic ring loses its lubricant film. Corner radius 0.2 mm maximum, with no continuous helical tool mark, since one spiral mark around the groove floor is a leak path. Finish the groove with a radial plunge or a single face cut, never a turning pass along the axis.

Thread concentricity and the sensing face

The thread mounts the sensor, so the thread axis is the reference for everything else. An M12x1 thread milled on a turn-mill holds 0.03 mm concentricity to the bore it was cut from. A die-cut thread holds 0.08 to 0.12 mm with error that is not repeatable, so the face tilts differently on every part.

An inductive sensor's range is set by the coil-to-target gap. Face runout of 0.1 mm at a 6 mm radius tilts the face about 1 degree, which on a 4 mm range with 10 percent repeatability costs roughly 0.05 mm of effective range. Small, but it moves the switching point differently on each unit. Specify 0.05 mm thread-to-bore concentricity and 0.05 mm face perpendicularity, then leave the method open.

Brass, 303, 316L or POM

Material follows the media, not the price list.

Brass is the default for general industrial sensors: three times faster to machine than 316L and a cleaner thread flank. Move to 316L for washdown, marine air or food lines. POM suits capacitive and magnetic sensors where a metal body would distort the field, but only below about 10 bar.

What to put on the drawing

State the environment before the geometry: media, pressure, temperature range, static or cycled joint. Then the cord diameter and squeeze, not a groove width. Call the thread class explicitly, 6g external and 6H internal, and say whether threads are masked during plating — electroless nickel at 8 µm adds 0.008 mm per surface and closes a 6g thread until it will not start by hand.

Name the face the seal seats on and how flat it must be. A 0.05 mm flatness callout on a 16 mm shoulder is cheap on a lathe, unrecoverable after bad masking. Send the O-ring spec and mating bracket with the housing and the DFM notes get useful; you can send both files here for a quote in 24 hours.

Frequently asked questions

How tight should the O-ring groove depth be?

Hold ±0.05 mm for any cord from 1.5 to 3 mm. That keeps squeeze inside 22 to 27 percent even when the molded ring varies ±0.08 mm on cord. At ±0.1 mm it still seals under pressure but may weep at 0.5 bar, so never on a vacuum part.

Can a 316L body still reach IP68 after machining?

Yes, if the cap bore and connector seat are cut in the same setup as the thread and finished to Ra 1.6 µm. The usual failure is not the alloy but 0.02 mm of taper in the cap bore, which makes the bond line vary. Pressure-test the first article at 1.5 times rated before releasing the batch.

Why does potting crack at -40 °C?

Expansion mismatch. Epoxy runs 40 to 60 µm/m·K against brass at 19 and aluminum at 23. A rigid 25 mm potted column in a brass shell develops enough shear at the bond line to crack the fillet. Use a flexible polyurethane compound, keep pot depth under 20 mm, or leave a 0.5 mm compliant layer.

What thread class should I call out?

6g external and 6H internal. Avoid 6e or a custom limit unless the mating part is also controlled, because plating thickness alone shifts a thread by half a class. On a plated body, specify the pre-plate size and let the shop compensate.

Related reading

• Robot End Effector Machining: Where Gripper Accuracy Actually Lives2026-09-19
• Liquid Cooling Plate Machining: Channel Depth, Fin Thickness and Flatness2026-09-25
• Vacuum Chamber Machining: Billet Quality, Sealing Faces and Virtual Leaks2026-09-20