The single number that predicts whether a seal is still sealing in year five. What it measures, how to read a percentage, what drives it, and how to specify it so the answer means something.
Overview
Compression set is how much of its original thickness a rubber part fails to recover after being held squashed. Squeeze a specimen by a fixed amount, hold it at temperature for a fixed time, release it, and measure what came back. The number is reported as a percentage of the deflection that did not recover, so lower is better: zero percent is a part that sprang all the way back, and 100 percent is a part that stayed exactly where it was crushed.
It matters because almost every static seal works by being compressed and pushing back. The clamp load is set at assembly and never renewed. Once the rubber stops pushing back, the joint leaks, and nothing about the incoming inspection would have predicted it, because the part passed hardness and tensile on day one.
How to read the number
| Result | What it means |
|---|---|
| Under 15 percent | Excellent recovery. Typical of a peroxide-cured EPDM or a good FKM at its rated temperature. |
| 15–25 percent | Normal for a well-specified sealing compound in general service. |
| 25–40 percent | Acceptable for a static gasket in a bolted joint that gets re-torqued, marginal for a long-life seal. |
| Over 40 percent | The part has largely stopped pushing back. Expect to see it as a leak eventually. |
A percentage on its own is meaningless without the conditions attached to it. Twenty percent after 22 hours at 70 °C and 20 percent after 1,000 hours at 150 °C are not the same material behaviour, and comparing two compounds tested under different conditions tells you nothing. Time, temperature, deflection, and specimen type all belong with the number.
The method: ASTM D395
- Method B, constant deflection, is the common one: the specimen is held at a fixed compression, usually 25 percent, in a jig
- Method A, constant force, holds a load instead, which is closer to a spring-loaded application than a bolted one
- Standard conditions are often 22 or 70 hours at a temperature chosen for the service, with longer pulls used when the question is service life
- Solid discs and plied discs give different results, so the specimen type is part of the spec
- Recovery is measured after a fixed cool-down, because rubber keeps recovering for a while after release
We run D395 in our own ISO/IEC 17025:2017 accredited lab, on the same site that mixed the compound and molded the part. A sulfur-versus-peroxide comparison takes a week here rather than two months.