Guide
What do burst, proof and cycling tests mean for the wall and thread design of a machined fitting?
Burst, proof and cycling tests are three different checks on a pressure fitting or coupling: burst finds the pressure at which it fails, proof shows it holds a set pressure without leaking, and cycling shows it survives repeated use. Each one punishes a different machined feature, and none of them can be guaranteed by a machine shop from the drawing alone, so the test plan has to be written into the order.
By Saola Manufacturing engineering team · Published · Updated
On this page
What does each of burst, proof and cycle testing prove?
Burst tests the margin to failure, proof tests that the part holds a stated pressure with no leakage, and cycling tests that it still holds after repeated connections or pressure events. They answer different questions, so a pass on one says little about the others.
The ASHRAE TC 9.9 white paper defines burst pressure for a fluid coupling as the minimum internal pressure for catastrophic failure, and notes that a common failure mode at burst is elastomeric seal extrusion (ASHRAE TC 9.9, Water-Cooled Servers, 2019). The point is that a burst failure often starts at the seal, not in the wall, which pulls the gland and the clearance next to it into the picture.
The Open Compute Project quick disconnect specification shows how the three test types fit together in one coupling program. It sets a minimum burst pressure, runs a hydrostatic hold with a no-visible-leakage criterion, and then repeats the hydrostatic hold after mechanical cycling and after operational temperature cycling (OCP UQD v2.0.0, section 9). Burst is a destructive design check; the hold and cycle tests are what a production lot could be asked to repeat in part.
What burst margin do published sources use?
There is no single factor; ASHRAE says typical burst limits for fluid couplings may be 2x to 5x the operating pressure. The OCP coupling document uses a minimum burst of 300 psi(g) against a 100 psi(g) maximum operating pressure, which is a ratio of 3 (derived: 300 / 100).
ASHRAE's survey table of typical coupling characteristics lists a minimum burst of 35 bar for cabinet, chassis and drawer couplings with a 10 bar maximum operating pressure, and 17 bar against 5 bar for blade couplings (ASHRAE TC 9.9, table 6). Dividing gives ratios of 3.5 and 3.4 (derived: 35 / 10 and 17 / 5), inside the 2x to 5x range. The same paper reports that the test multiple for liquid-filled components in the IEC safety standard moved from 5x to 3x under normal conditions and 2x under single-fault conditions in the third edition, while the piping code ASME B31.3 tests at 1.5x design pressure.
Treat these as published reference points for coupling and cooling-loop work, not as a safety factor for your fitting. The right margin depends on the fluid, the consequence of failure, the applicable code and whoever owns the design. Do not take a multiple from this guide into a drawing note; take it from the standard or the customer specification that governs your product.
| Test | What it exercises | Published values and basis |
|---|---|---|
| Burst | Wall, closure joints, thread roots, seal extrusion | Typical limits may be 2x to 5x operating pressure (ASHRAE); OCP minimum burst 300 psi(g) against 100 psi(g) maximum operating, a ratio of 3 (derived: 300 / 100) |
| Proof or hydrostatic hold | Sealing faces, glands, thread seals | OCP: pressurize to 100 psi(g), hold at least 3 minutes, no visible leakage; its burst test runs to failure or to 3x the minimum proof pressure, whichever is lower |
| Mechanical cycling | Wear on sealing surfaces, latch, edges that touch the seal | OCP: minimum five thousand cycles at 40 psi(g), then a hydrostatic hold; ASHRAE survey shows typical connection cycles of 100 to 1000 by location |
| Seal washout under flow | Gland edge geometry and extrusion gap | OCP: 20 cycles in each flow direction at twice the rated flow |
| Temperature cycling | Differential growth and seal behavior | OCP: operating test cycles between 5°C and 65°C with 72 hours at each extreme, then a hydrostatic hold |
Which machined features fail first in a burst test?
The seal clearance, the thread root and the thinnest wall, usually in that order of suspicion. Seal extrusion through the gap next to a gland is the failure mode ASHRAE names, so the diameters that bound the gland are the first dimensions to control.
A diametral clearance beside an elastomer seal is an extrusion gap, and the machined tolerances on the two mating diameters set it. The OCP document shows the design attention this gets: one socket dimension is given as a minimum-to-maximum range for the larger sizes to reduce the extrusion gap and mitigate seal washout (OCP UQD v2.0.0, table 1 footnote). On your drawing, tolerance the pair of diameters bounding the gland so that the worst-case gap is acceptable, and read the O-ring groove guide for gland dimensions.
The thread root is a stress concentration. A sharp root radius, a thread that runs out at a thin section, or a thread cut through a cross hole will start a crack before the wall yields, so use a thread relief and a defined root, as described in the thread relief groove guide. The thinnest wall is often not where you expect: a groove cut into a bore leaves a thinner section under the groove, and a cross hole removes metal at the highest stress. Check the remaining wall at the groove, the cross hole and the thread run-out, not only the nominal wall.
What does cycling do to a machined coupling?
It wears the surfaces that touch the seal and the latch, and it exposes any edge sharp enough to cut the seal. The features at risk are chamfers and radii at seal entry, sealing-surface roughness, and mating hardness of the parts that slide.
Each connection drags the seal over the plug end, so the entry chamfer or radius and the surface texture decide how long the seal lasts. The OCP appendix says a version 1 plug end with minimal or no chamfer may damage the seals of both version 1 and version 2 sockets, and its plug detail shows a minimum radius or equivalent chamfer at the end (OCP UQD v2.0.0, section 15.1 and figure 8). Its material table requires the stainless grades 303, 304, 304L, 316 or 316L at the plug tip and the cone and guide, in order to keep similar hardness on parts in forceful contact.
Stainless parts that slide or thread against each other can gall; the thread galling guide covers the causes. The seal washout test adds a second cycling load, 20 connections in each direction at twice the rated flow, aimed at seals dislodged from their gland (OCP UQD v2.0.0, section 9.1.5). That makes gland edge breaks, and burr-free intersections near the seal, part of the cycle-life drawing and not a cosmetic note.
Which standard covers which connection?
ISO 18869 covers test methods for quick-action couplings, and ISO 19879 covers test methods for metallic tube connections, stud ends and flange connections. They are separate documents with independent test methods, so name the one that matches your part.
The ISO 18869:2017 record describes methods for testing and evaluating the performance of quick-action couplings, says the methods are independent and chosen by the user, and notes that the connector standard states which tests to run (ISO 18869:2017 catalog record). The record lists a revision in development, so check the edition you are citing. ISO 19879:2021 specifies uniform methods for testing metallic tube connections and stud ends for ports and flange connections, and it states that quick-action couplings fall under ISO 18869 instead (ISO 19879:2021 catalog record).
The OCP document uses ISO 18869 for rates of pressure rise and sample size, with at least five samples per test, and uses ASTM E493 for helium leak tests. If your product is a stud end or tube fitting, ask the design owner which method applies, and cite the standard, edition year and test number on the drawing or test plan. These ISO documents are paid standards, so this guide describes their scope and does not reproduce their values.
What can a build-to-print supplier guarantee, and what stays with the design owner?
A supplier can guarantee that the metal parts meet the drawing and report the evidence; it cannot guarantee burst margin or cycle life, which depend on the design, the seal and the assembly. Those are validated by the design owner on assembled samples.
For liquid-cooled IT equipment, ASHRAE says the IT manufacturer is responsible for pressure testing the assembly it ships. For a machined fitting, the supplier's evidence is the dimensional report, the thread gauging record, the material record and the visual and burr inspection. Burst is destructive, so it is run on validation samples, and production lots are typically accepted on dimensions and a non-destructive leak or proof check agreed in advance.
Saola machines fitting and coupling bodies by CNC turning and CNC milling at its lead plant in Vietnam, and arranges surface treatments through partner plants it manages. It does not set your pressure rating. Where the buyer or an independent lab runs burst and cycle tests, Saola supplies the traceable samples and the measurement data to compare them with. The build-to-print responsibility guide explains why the line is drawn there.
How should the test plan go into the RFQ and drawing?
List each test, who runs it, how many samples, with which fluid and acceptance criterion, and which standard and edition define it. Ask the supplier to quote the extra validation samples as a separate line, with their lot and material traceability.
For each test, write the method reference (for example ISO 18869 for a coupling), the sample size, the rate of pressure rise, the fluid and the pass criterion. The OCP template shows the level of detail: each test has an identifier, pass or fail per sample for plug, socket and mated pair, and a unit of measure (OCP UQD v2.0.0, appendix 15.3). Say which tests happen at the supplier, which at your site or a lab, and whether the supplier witnesses them.
On the drawing, tie each test to the features it exercises: the gland diameters and extrusion gap, the thread class and root radius, the minimum wall at grooves and cross holes, and the burr and edge notes. Add the first-article requirement, so the first samples are measured before they are pressurized; the first article inspection guide explains the report. If a requirement is not written, assume it is not part of the quote.
Frequently asked
Is proof pressure the same as test pressure?
Not necessarily. Proof is a stated pressure the part must hold without leaking or permanent deformation, while a test pressure may be a leak check at a lower value. The OCP document uses a hydrostatic hold and refers to a minimum proof pressure when capping the burst test, so define the terms in your own test plan.
Can I specify a safety factor on the drawing?
Only the factor your governing standard or customer specification requires. Published references such as ASHRAE give typical ranges for coupling burst margin, not a requirement for your product.
Does a higher burst rating mean a longer cycle life?
No. Cycle life depends on seal material, surface finish, edge condition and latch wear, which are separate from wall strength. Run the cycle test on the assembled part.
How many samples does a burst or cycle test need?
The OCP document, citing ISO 18869, uses at least five samples per test for its property tests. Your plan should state the number and why.
Can a CNC supplier run the burst test?
Only if it has the equipment and you agree to it in writing. Many buyers run validation tests themselves or at a lab, and the supplier provides traceable samples and dimensional data.
Sources
- Water-Cooled Servers: Common Designs, Components, and Processes, ASHRAE TC 9.9 white paper, ASHRAE, 2019 (errata October 2019)
- Open Compute Project, Universal Quick Disconnect (UQD) Specification, version 2.0.0, Open Compute Project, document dated September 22, 2026
- ISO 19879:2021, Metallic tube connections for fluid power and general use: test methods, ISO catalog record via Genorma, published March 2021
- ISO 18869:2017, Hydraulic fluid power: test methods for couplings actuated with or without tools, ISO catalog record via Genorma, published July 2017