Buying O-Rings and Seals for HVAC and Pump Assemblies in China: Squeeze, Groove Fill and the Material Mix-Up

I spent a morning at a Hefei factory last spring watching a body shop handle a leaking O-ring problem that had been running for eleven months. They had changed the O-ring supplier three times. They had changed the groove dimensions twice. They had run silicone, then EPDM, then silicone again. It still leaked, but only on the units that sat on the dock over the weekend.

The actual cause took about forty minutes to find once someone checked the temperature of the water jacket, the coolant chemistry, and the fact that the so-called EPDM they had bought twice was a cheaper peroxide-cured grade being sold as the same thing. The O-ring was never the problem. It was the last part in a chain, and it was the only part anyone could easily swap.

That pattern is, in my experience, how most leaks in HVAC and pump assemblies actually work. So before writing a spec, it helps to understand what the groove is doing.

Why an O-ring seals at all

An O-ring does not work by being squeezed into the gap. It works because it is compressed, and that compression makes the rubber push outward and inward against the groove and the mating surface, sealing at contact pressure. The system pressure then adds to that. That is the whole mechanism, and it has three requirements: right squeeze, right groove fill, right surface finish.

Squeeze is (free cross-section − groove depth) / free cross-section. A static radial seal typically wants 15–25% on a fixed face, and 12–20% on a face where thermal cycling moves the parts. Below about 10%, you are relying on system pressure to do the sealing, which works right up until the pressure drops, and then you leak on a weekend when nobody is looking. Above about 30%, the rubber takes a permanent set and fails as a leak a year later.

Groove fill is the other number nobody asks about. The groove cross-section has to be larger than the O-ring cross-section so the rubber has somewhere to go when it is squeezed and when it swells. Fill ratio of 70–85% is the usual band. If you squeeze a 3.53 mm cord into a groove sized for 3.0 mm, the rubber has no room, the squeeze becomes uneven, and it will find a way out.

Surface finish is the third leg. A groove machined with a rough turning insert leaves a spiral groove that is a leak path no O-ring can seal, no matter how soft. For static seals I want roughly Ra 1.6 µm or better in the groove, and no visible tool marks across the sealing direction. Plunge or radial face grooves, not a thread that happens to end there.

Material is a chemistry decision, and China has a bait-and-switch problem

NBR is the default for water, air and general hydraulic oil up to about 100 °C. EPDM is for hot water, steam, glycol and anything exposed to weather and ozone. FKM (Viton-type) is for fuel, aggressive solvents and heat to 200 °C. HNBR for a bit more heat and better mechanical strength than NBR. Silicone is for very wide temperature range and low force — and it is soft, poor in water with pressure, and easy to tear.

None of these are interchangeable, and this is where the Chinese market gets confusing, because the trade names are not standardised. Two shipments of "EPDM 70" from two different factories can be EPDM and EPDM, or EPDM and a blend with recycled rubber, or EPDM cured with the wrong system for your application.

The concrete check is the datasheet plus the compound number, not the material name. Ask for the compound code, the cure system, the hardness with tolerance, the compression set at your service temperature (70 hours at 100 °C is the usual basis for NBR/EPDM, 70 hours at 175 °C for FKM), and the compression set value in percent. Anything under about 20–25% is a real, well-formulated compound. Values above 35% mean you will be replacing seals.

Also ask what the material actually is under a lab check. Silicone passes are easy. EPDM grade verification is harder, and a TGA or FTIR plus a solubility check is what an independent lab does. I have paid for that test twice in six years and both times it saved me more than it cost.

The counterintuitive one: temperature cycling is worse than high temperature

If you have a seal that sees 90 °C constantly, you can pick a material for 90 °C and be done. If you have a seal that sits at 20 °C, warms to 95 °C during a cycle, and comes back down, you have two problems. The differential expansion between the metal housing and the rubber creates a pumping action that can draw fluid past the seal. And the repeated movement of the rubber in the groove abrades it.

Two things help. A larger and more consistent squeeze so the seal keeps contact through the thermal excursion, and a groove designed so the O-ring cannot roll or extrude. On the extrusion side, the data point to look for is the gap. For a 70 Shore A seal at 10 bar and 100 °C, the diametral clearance limit is roughly 0.2 mm. Go past it and the rubber extrudes into the gap and the joint starts weeping. Back-up rings are the professional answer, and also a reliable way to spot a serious factory: they will suggest them.

Assembly is where most of the leaks are born

The majority of field leaks I have investigated were installed correctly on paper and badly in practice. The three killers, in order of frequency:

Nicked or cut seal during installation, usually on a thread or a sharp housing edge. A proper chamfer with a radius and an installation cone fixes this. Cheaper than a recall.
Dry assembly. An O-ring pushed into a dry groove scuffs itself and never recovers. Specify a compatible lubricant — and "compatible" matters, because a silicone grease on a silicone seal is fine but a silicone grease on a part that will be painted is a nightmare.
Over-torqued or unevenly tightened flanges. Two bolts tightened one after the other will leave the seal differentially compressed and a leak at the far side.

These are not supplier defects. But if the assembly is done at the factory, they are your defects as the buyer, because it is your name on the unit.

What goes on the purchase order

A workable PO line for a static seal, from a buyer who has been burned:

Compound code and cure system stated, not just "EPDM"
Hardness 70 Shore A ±5, compression set <25% at service temperature, 70 h
Dimensions to the standard you are actually using — AS568, metric ISO 3601, or your own drawing with groove dims — and if custom, the groove detail supplied with the drawing
Groove finish Ra ≤1.6 µm, no spiral tool marks, chamfer per drawing
Dimensional check per lot: cross-section and ID/OD on a sample, with a report
Batch and cure date on the bag, shelf life stated, storage conditions stated
Mould ownership, if a custom profile, in writing

The shelf-life line matters more than buyers expect. Rubber ages in storage, and a factory shipping you the bottom of a drum that has been sitting for two years is giving you a seal with reduced compression set from day one. Get the manufacturing date, and reject anything over a year old for critical joints.

One more thing I keep coming back to: when a leak appears, resist the urge to change the seal supplier first. Measure the groove, check the material against the drawing, check the surface finish, and look at the assembly. In eleven months at that Hefei factory they had done everything except the three things that would have found the cause in an afternoon.