# Is it possible to use radial seals on non-circular cross sections?

• berle
berle
TL;DR Summary
Can you use radial seals on non-circular cross section (rectangle with filleted corners for instance) for high pressure applications (subsea).
Hi everyone,

I am in the process of designing a watertight enclosure for 300m deep applications and would like to explore options of using radial o-ring seals in place of face-seals due to space-savings. However, i am unable to find any information anywhere on using radial seals on non-circular cross-sections. Basically, the shape is a rectangle with LARGE radius corners.

Any inputs or ideas on where to find information on the subject would be appreciated :).

Thanks!

Welcome to PF.

berle said:
However, i am unable to find any information anywhere on using radial seals on non-circular cross-sections.
I think you need to produce a couple of diagrams that explain how the axial face seal was done, and how the radial o-ring seal will be implemented.

Will the groove for the o-ring be cut in the plug or the aperture?
What will stop the plug being pushed in, a taper or a step?
How will you machine the groove for the o-ring?

Of course, I have attached an image of the idea for the end "plug".
I have also thought a lot about how the grooves will be machined, and I suppose it would at least be machinable on a 5-axis mill.
The only real reason to do it this way is to eliminate the need for thicker walls at the sealing interface to be able to have space for both bolts and o-ring. Otherwise a face-seal would be my go-to solution.

Last edited by a moderator:
I can think of no reason (theoretically) that you couldn't use O-rings in a static application with the 'gentled' rectangle that you describe. The issue is:
Any 'differences' along the O-ring (from the POV of the O-ring) in terms of groove geometry or clearance with the mating surface will cause local tension (in the O-ring) to vary. Some of that is normal, but a circular shape allows a properly lubricated O-ring to 'slide' and equalize tension; a rectangular shape probably doesn't allow as much of that. Tension differences cause thinning, and reduce the size of the O-ring. I wouldn't be afraid to try it, but I'd make (extra) sure that the tolerances, surface finishes, and assembly procedure (lubrication, cleanliness, tension distribution) were properly executed.

Disclaimer: I may be a little over-the-top - one of my formative experiences was screwing up an O-ring design in an unusually expensive (then) material.

The problem I see with a radial seal, is extrusion of the o-rings.

With a face seal, the gap between the two parts is closed by pressure, then the o-ring is pushed up against the junction by fluid differential pressure, which seals the joint.

With your radial system, the gap between the two parts is not closed by the pressure, so where the o-ring is pushed against the junction, there will be a gap, sufficient to cut and extrude the o-rings. Damage to the o-rings during assembly is also more likely. The better the fit, the better the seal and the more difficult will be the assembly.

If I had to use radial seals, I would employ a tapered plug, that closes the gap before the flanges contact. That would be difficult to manufacture if it was not conical, which could be lapped. That may explain why non-circular cross-sections are not used with radial seals.

Radial seals with multiple o-rings are used on circular plugs, such as the rod ends of hydraulic cylinders, but those are circular, precision turned, employ tapers at the junctions to aid assembly, and are subjected to internal pressure.

Dullard
It should be possible to do the concept in Post #3. You would need tight tolerances and a close fit between the two parts in order to get good compression on the O-rings and prevent the extrusion mentioned by @Baluncore in Post #5. You would also need a taper in the mating part as indicated in the crude sketch below. And installing the part would still require some sort of fixture or careful handling to keep the O-rings fully in their slots during installation.

The machining could be done with a 3 axis mill if the machinist used a key slot cutter.

Dullard and berkeman

Longer answer: as previously mentioned by the other folks here, it’ll be a little tricky to get it to work right. I advocate using a healthy dose of lubricant during both installing the o-ring and closing out the assembly. If it’s not a chemistry issue, I strongly recommend using petroleum jelly as your lubricant. I also encourage you to take your time with getting the o-rings installed, and to not be afraid of wiggling them around in the grooves a bit to get them settled evenly. It won’t eliminate the risk of them getting extruded or twisted, but it will certainly reduce the chance. Again, the petroleum jelly will be helpful for that, as it’s tacky/grippy enough to help hold the rings in place.

On a somewhat tangential note: is there a particular reason why you’re opting for the box shape instead of a cylinder? Space constraints or something?

Dullard said:
I can think of no reason (theoretically) that you couldn't use O-rings in a static application with the 'gentled' rectangle that you describe. The issue is:
Any 'differences' along the O-ring (from the POV of the O-ring) in terms of groove geometry or clearance with the mating surface will cause local tension (in the O-ring) to vary. Some of that is normal, but a circular shape allows a properly lubricated O-ring to 'slide' and equalize tension; a rectangular shape probably doesn't allow as much of that. Tension differences cause thinning, and reduce the size of the O-ring. I wouldn't be afraid to try it, but I'd make (extra) sure that the tolerances, surface finishes, and assembly procedure (lubrication, cleanliness, tension distribution) were properly executed.

Disclaimer: I may be a little over-the-top - one of my formative experiences was screwing up an O-ring design in an unusually expensive (then) material.
https://en.wikipedia.org/wiki/Percy_Williams_Bridgman

Could be "pinch-off" failure of softer O-ring compounds; see Bridgman.

• Mechanical Engineering
Replies
1
Views
789
• Mechanical Engineering
Replies
1
Views
2K
• Mechanical Engineering
Replies
15
Views
815
• Electrical Engineering
Replies
11
Views
3K
• Electrical Engineering
Replies
14
Views
1K
• Classical Physics
Replies
13
Views
1K
• Atomic and Condensed Matter
Replies
1
Views
2K
Replies
1
Views
1K
• Nuclear Engineering
Replies
2
Views
2K
• Mechanical Engineering
Replies
1
Views
2K