Home / Valve Test Benches / Hydrostatic test procedure
The valve hydrostatic test procedure: eight operation steps, in order
Most failed valve tests are not lost at the pump. They are lost at the sealing plate, at the vent, or at the gauge range — three of the eight steps below, and all three are decided before the pressure goes up. This is the sequence we write into every operation manual: what has to be settled before the valve is clamped, what happens under pressure, and what has to leave the bench with the valve for the test to be worth anything afterwards.
Quick answer
| Steps | 8 — six of them settled before the pump starts |
|---|---|
| Test pressure | Shell 1.5 × PN, seat 1.1 × PN |
| Hold time | By size and class; 1–2 minutes for judgement in most bands |
| 0–40 MPa | Two gauges, one per range; pressure set by valve class |
| During the hold | Unloading pressure holding — pump unloaded, cavity locked |
| Safety | No motion command accepted while pressurised |
| Output | Per-run time–pressure curve, exported |
The eight valve pressure testing machine operation steps
Written as a table, the whole procedure fits on one page — which is the point. A step that cannot be stated in a line is a step that will be done differently by every shift.
| # | Step | What is decided or verified | Where it usually goes wrong |
|---|---|---|---|
| 1 | Read the valve, fix the test pressure | Type, DN, class → shell and seat pressures, hold time | Test pressure taken from the machine's maximum instead of the valve class |
| 2 | Change the sealing plates | Plate matched to flange standard, class and DN | A plate from the last job that “almost fits” |
| 3 | Load and clamp | Clamp force and jaw/ram position; no bolt set | Valve not seated flat on the plate before clamping |
| 4 | Fill and vent | Both cavities filled, high point vented clear | Air left in the cavity — the classic bad curve |
| 5 | Pressurise | Setpoint on the contact gauge; correct gauge range | Low-pressure test read on the bottom of a 40 MPa dial |
| 6 | Hold | Unloading pressure holding; interlock armed | Operator tries to unclamp before release to zero |
| 7 | Judge the four tests | Shell, seat, back seat, overall — one clamping | Pressure drop attributed to the valve when it was the fixture |
| 8 | Depressurise and export | Release to zero, unload, export the curve | Test passed but no record leaves the plant |
Steps 1 to 4 are decisions. Steps 5 and 6 are the machine's job. Steps 7 and 8 are judgement and evidence. If a plant is having trouble with repeatability, the cause is almost always in the first group, not the second.
Before the pump starts: steps 1 to 4
Step 1 — Read the valve, not the gauge
Decided here: two pressures and a hold time.
The test pressure is a property of the valve, not of the machine. Valve type, DN and pressure class come off the nameplate and the purchase specification; the pressures are calculated from them. For a PN-rated valve the shell strength test runs at 1.5 × PN and the seat test at 1.1 × PN.
This is also the step where the bench's own limit has to be read honestly. A bench quoted as reaching up to 40 MPa delivers that figure with two gauges — one gauge per range — and the pressure actually used is set by the valve class. Above DN350 the working band drops to the class of the valves in that size, which is why large-bore benches are quoted at 10 or 16 MPa rather than at the small-bore maximum. Quoting the maximum against a large bore is how a test specification becomes unmeetable. The full class-by-class arithmetic is in the hydrostatic test pressure chart.
Step 2 — Change the sealing plates before you change the valve
Decided here: whether the test measures the valve or the fixture.
Sealing plates are matched to the flange standard, the pressure class and the DN. A plate left over from the previous job that “almost fits” is the most common single cause of a pressure drop that gets blamed on the valve. The check takes a minute:
- Sealing face clean — no scale, no dried compound, no swarf.
- No nick or score across the sealing line (a scratch on the outside edge is harmless; one crossing the seal is not).
- Seal element not compressed flat by a previous over-clamp.
- Plate seating square in its recess, not tilted in the bore.
On a bench with automatic clamping the plate is changed by hand and the clamp does the rest — the plate is still the part that decides what the test is measuring. The three fixture families and the body styles each one suits are set out in the clamping methods reference.
Step 3 — Load and clamp: no bolts to tighten
Decided here: clamp force, and whether the body sits flat.
The valve goes on and the clamp closes. On a vertical bench the body is lowered onto the sealing plate and held by the clamp — on the large-bore SYJ-600M, a 126-tonne vertical clamp on four jaws covers DN350 to DN600. On a horizontal bench the body is rolled in and clamped from the ends, which is the reason the WBD-600 exists: a 200 kg body does not have to be turned over between faces.
What disappears at this step is the flange bolt set. Automatic clamping replaces the bolt circle a manual blind-plate setup needs, so a DN600 valve is a one-person job with a hoist rather than a two-person job with a spanner. What does not disappear is the seating check: a body clamped while sitting on a bit of grit is clamped crooked, and it will leak at the plate, not at the seat.
Step 4 — Fill and vent: this is where tests are lost
Decided here: whether the pressure reading means anything.
Water enters through the blind plates — top and bottom on a vertical bench, left and right on a horizontal one — and the high point is vented until water runs clear with no sputter. Air is compressible and water is not. A cavity that is only partly filled climbs unevenly, makes the gauge lag the pump, and on release makes the pressure look like it is holding while the cavity is still filling.
Dual-port filling is what makes step 4 worth doing properly. Because each plate has its own water path, the upper and lower cavity are filled and tested in sequence without unclamping the valve — which is the practical basis for running four tests on one setup in step 7.
Under pressure: steps 5 and 6
Step 5 — Pressurise to the setpoint
Decided here: which range the reading comes from.
Pressure is set on the contact pressure gauge at the console before the run, and the pump stops at the setpoint instead of being watched up by eye. Setting it at the console rather than at a valve behind the machine matters for one reason: the operator who sets the pressure is the operator who will read it, and he should be looking at the same dial for both.
On a bench covering 0–40 MPa the range is split across two gauges — a low-range gauge for the low-pressure work and a high-range gauge for the high-pressure work, each read on its own scale. One gauge stretched across the full range would put every low-pressure test in the bottom tenth of the dial, where reading error is largest, and that bottom tenth is where most valve testing actually happens.
Step 6 — Hold: unloading pressure holding
Decided here: nothing. This is the machine's step.
At the setpoint the pump is unloaded — the oil returns to the tank instead of being held against a closed circuit — and the valve cavity stays locked by a check valve with the gauge live. The short version used on the floor is that the machine rests while the pressure stays. Two consequences matter commercially: oil temperature does not climb through a shift of continuous testing, and the hold does not consume pump life. A large oil tank adds cooling margin on top of that, which is what keeps a bench that runs all day from running hot by mid-afternoon.
While the cavity is pressurised, the controller refuses every motion command. Buttons do not act; the clamp will not open; the pressure has to be released to zero first. This is the step an operator is most likely to try to short-cut, and it is the step the interlock exists for — not for convenience, but because the alternative is opening a clamp on a live cavity.
Judgement and handover: steps 7 and 8
Step 7 — Judge the four tests
Decided here: which of the four results is a fail, and why.
Shell strength, seat tightness, back-seat tightness and overall tightness are four different tests on the same valve, and on a bench with dual-port filling they run in sequence without the valve coming off: shell first, then the seat tests, then the overall test. Four verdicts, one clamping — no re-clamping between test types, which is the difference between a four-test procedure and four separate setups.
What is being judged is the shape of the hold rather than a single number at its end. A flat line is a pass; a step drop in the first seconds usually means the fixture moved and the run should be repeated before the valve is condemned; a slow decay through the whole hold is leakage and belongs to the valve. The four shapes and what each one means are set out in the records and pressure curve reference, and the acceptance side — what a rate class actually permits — is in the API 598 and ISO 5208 leakage class reference.
Step 8 — Depressurise, unload, export
Decided here: whether the test still exists in eighteen months.
Pressure is released to zero, the clamp opens, the valve comes off, the plates go back on the rack. The run is already stored with its curve; the export goes to a USB stick and into the plant archive, where it is found later by valve serial rather than by scrolling through a month of runs.
A test that is not exported is a test that will be argued about. The curve is the only artefact of the whole eight-step procedure that survives the valve leaving the building, and it is what an incoming-inspection department or a third-party surveyor will ask for first.
The numbers this procedure runs on
Two pressures and a hold time, both taken from the valve class rather than from the bench. The table below is the arithmetic for the classes most often tested on a flanged valve; the complete PN10–PN160 table, with the hold times by size, is in the pressure chart.
| Valve class | Shell test (1.5 × PN) | Seat test (1.1 × PN) | Typical bench |
|---|---|---|---|
| PN16 | 2.4 MPa | 1.76 MPa | Any bench in the range; low-pressure gauge |
| PN25 | 3.75 MPa | 2.75 MPa | Any bench in the range; low-pressure gauge |
| PN40 | 6.0 MPa | 4.4 MPa | Vertical bench, DN15–DN300 band |
| PN63 | 9.45 MPa | 6.93 MPa | Vertical bench, small to medium bore |
| PN100 | 15.0 MPa | 11.0 MPa | Small bore only; high-pressure gauge |
Pressure is always stated against a bore and a class. A bench advertised at up to 40 MPa reaches that figure with two gauges, one per range, and the pressure used is set by the valve class — not by what the pump can do. Above DN350 the practical band is the class of the valves in that size, which is why large-bore benches are quoted at 10 or 16 MPa.
Where the procedure breaks down
Five failure modes account for most of the bad tests we are asked about, and only one of them is a machine fault:
- Plate mismatch (step 2). The pressure drop is at the plate, not the seat, and the valve is scrapped for nothing. Cure: plate checks written into the work instruction, not left to memory.
- Air in the cavity (step 4). The ramp is ragged, the gauge lags, and the hold looks unstable on a perfectly good valve. Cure: a venting step with a stated end condition — water clear, no sputter.
- Wrong gauge range (step 5). A low-pressure result is read at the bottom of a high-range dial; the verdict is technically recorded and practically meaningless. Cure: two gauges, one per range, and a rule about which one is read.
- Unclamping early (step 6). Prevented mechanically by the interlock on our benches, but still attempted; the interlock is what makes the procedure safe rather than merely written down.
- No record (step 8). The test happened, nobody can prove it, and a field failure becomes an argument. Cure: export as part of the step, before the valve leaves the bench.
What this procedure cannot promise
A bench can hold a pressure, hold it steadily, and record what happened while it did. It cannot promise that a particular valve will not drop pressure — that depends on the machining of the sealing faces and the condition of the seat, both of which are upstream of the test. A procedure that says so is more useful than one that does not, because it tells the operator where to look when a curve misbehaves: at the plate, at the vent, at the gauge range, and only then at the valve.
The same honesty applies to the numbers. Hold time follows the standard for the size and class rather than a habit; a longer hold does not prove more, and on a sealing face that is not flat it only shows the drop more clearly.
Practical questions around the procedure — how long one valve takes, whether a test certificate comes out per valve, which standard sets the hold times — are answered in the valve test bench FAQ.
Which bench for which part of the procedure
SYJ-150 · Small bore, high pressure
- DN15 – DN150, 30 t clamp
- Up to 40 MPa, delivered by two gauges
- Where the dual-gauge range earns its place
SYJ-600M · Large bore, vertical
- DN350 – DN600, 126 t four-jaw clamp
- Water to 10 MPa, set by valve class
- Loaded from above with a hoist
SYJ-600SQ · Water and air
- DN50 – DN600, 90 t
- 40 MPa water / 4 MPa air
- Dual-port filling, four tests in one clamping
DFS2-600 · Dual station
- Left DN50 – DN300 / right DN350 – DN600
- 0 – 7 MPa
- Small and large valves tested in parallel
For the sequence that surrounds this one — pressure, leakage class, layout, fixture, acceptance and records — start at the valve pressure test bench buyer guide, then the pressure chart, the leakage rate classes, the layout comparison, the clamping methods and the FAT checklist. Plants running gate valves in volume may also want the ZZC2-200 automatic gate valve test line, where the same eight steps run as a station sequence.
Frequently asked questions
What are the steps of a valve hydrostatic test?
Eight: fix the test pressure from the valve class, change the sealing plates to match the flange, load and clamp, fill and vent, pressurise to the setpoint, hold, judge the four tests, then depressurise and export the record. Six of the eight are decided before the pump starts, which is why a written procedure beats an experienced operator on the day the night shift runs the bench.
Can the shell test and the seat tests be run in one clamping?
Yes, on a bench with dual-port filling: water enters through both blind plates, so the upper and lower cavity are pressurised in sequence and shell strength, seat tightness, back-seat tightness and overall tightness are four tests on one setup. The valve does not come off between test types.
Why does trapped air ruin a hydrostatic test?
Air is compressible and water is not. A cavity that is only partly filled climbs unevenly, makes the gauge lag the pump, and on release makes the pressure appear to hold while the cavity is still filling. The curve produced that way cannot be defended, which is why venting is a step in its own right and not a detail of filling.
Why does a 0–40 MPa bench use two pressure gauges?
Because reading error is largest at the bottom of a dial. One gauge per range keeps low-pressure tests in the middle of their own scale instead of in the first tenth of a 40 MPa dial. The pressure used is still set by the valve class, not by the gauge maximum.
What happens if the clamp is opened during the hold?
It will not open. While the cavity is pressurised the controller refuses every motion command, and the pressure has to be released to zero before the clamp responds. The interlock exists for the moment an operator decides the test is taking too long.
How long should the hold last?
The hold follows valve size and pressure class under the applicable standard, not a fixed habit; for most sizes in the DN15–DN300 band the hold used for judgement is one to two minutes. Holding longer does not prove more — if the sealing face is not flat, a longer hold only shows the drop more clearly.
We send the procedure as a printable sheet with your machine, and in advance on request. Send your valve types, DN range and pressure class — the reply includes the model, the sealing plate set and the procedure, within 12 working hours.