Home / Valve Test Benches / Test records & pressure curves
The test record is the product — the curve is what makes it stand up
A valve that passed the works test carries no proof of it, except the record. When a customer's inspector, an incoming-inspection department or a third-party surveyor looks at that record, they rarely argue with the final number — they look at the shape behind it. This page sets out what a complete test record carries, how to read the four curve shapes a hold test can produce, and how the records stay traceable from the bench to your archive.
Quick answer
| What an inspector trusts | The time–pressure curve, not the final number alone |
|---|---|
| What a record must name | Valve identity, machine, medium, test type, pressures, hold, verdict, operator |
| What an instant drop means | Usually the fixture — re-clamp before condemning the valve |
| What a slow decay means | Real leakage — a valve question, not a machine question |
| Where records live | In the controller, exported over USB to your archive |
Why the curve, not the number
A single reading — "60 bar after 60 s: pass" — is easy to write and easy to dispute. Two valves can show the same end-of-hold pressure while their holds were nothing alike: one sat flat at the set pressure the whole time, the other dropped from a higher start and was caught by the acceptance margin at the last second. The final number cannot tell those tests apart. The curve can, at a glance, and that is why an inspector who has seen disputes asks for the curve before asking for anything else.
The curve also protects you. When a customer claims a field failure and traces it back to the works test, a flat curve on file is the end of the conversation; a report with only a number invites the argument to continue. Records are not paperwork produced after the test — on a properly specified bench they are a by-product of the test itself, produced whether or not anyone asks for them.
The eight fields a complete record carries
| # | Field | Why it is on the sheet |
|---|---|---|
| 1 | Valve identity — type, DN, pressure class, serial or batch number | Ties the record to the physical valve that ships; batch number covers production testing where per-valve serials are not used |
| 2 | Machine identity — model and serial number | Traceability runs both ways: a calibration question later must reach the exact machine |
| 3 | Test medium | Water or air — the leakage criteria differ, and a gas result is never interchangeable with a water result |
| 4 | Test type — shell, seat, back-seat | The same valve is tested at different pressures on different surfaces; the record must say which |
| 5 | Set pressure and hold time | For a PN-rated valve, the shell test at 1.5 × PN and the seat test at 1.1 × PN; hold time per the applicable standard |
| 6 | The time–pressure curve | The evidence itself — the whole hold, not a snapshot |
| 7 | Verdict against a stated criterion | Pass/alarm judged against the agreed acceptance factor or leakage rate class — a verdict without a criterion is an opinion |
| 8 | Operator and date | Accountability, and the entry point for any later investigation |
Field 7 is where records most often go soft. The criterion — test pressure, hold time, acceptance factor or leakage rate class — is agreed before the run, which is why the FAT checklist puts the acceptance settings on the agenda. What the leakage side of that criterion can legally claim is a separate subject: see API 598 and ISO 5208 leakage rate classes.
Reading the curve: four shapes and what each one means
Three of the four shapes below are routine; the skill is telling them apart fast, because two of them look like a failed valve and only one of them is one.
1. Flat hold — pass
Ramp up, small settling dip, then a flat line to the end of the hold.
The normal signature of a sound valve in a sound fixture. The verdict records the pass; the curve archives it.
2. Instant step drop — suspect the fixture first
Pressure collapses within seconds of reaching the setpoint.
A step that fast is rarely a seat leaking — it is usually mechanical: a sealing plate that slipped, a seal element that let go, a clamp that lost grip. The rule on the floor: re-clamp and run again before condemning the valve. If the second run repeats the step, look at the fixture and its sealing surfaces before writing the valve off.
3. Slow continuous decay — a valve question
Pressure creeps down steadily through the hold, never collapsing.
This is the shape of actual leakage — through the seat, or a hairline defect letting by. It is judged against the acceptance criterion: within the factor or rate class, it passes; below it, the alarm is real. This is also the shape a customer's inspector will zoom into, which is why the curve must be on the record, not summarised away.
4. Hunting around the setpoint — a machine tuning question
The trace oscillates instead of settling after the ramp.
Not a valve fault at all — it shows up later as nuisance alarms on good valves. It is settled during commissioning and checked at the FAT, which is why the FAT run is watched end to end rather than only at its verdict.
Traceability: the record must survive the machine
- Stored, not transcribed. Every run is stored in the controller with its time–pressure curve. Transcribing numbers onto a paper sheet by hand is where records lose their defence value — transcription is the one step nobody can audit.
- Exported off the machine. USB export of the run files, opened on an office computer before the courier collects the valve. The archive lives on your network; a record that exists only inside the machine is one hard-disk failure from disappearing. On the ZZC2-200 automatic line the record is printed at the station as well, for plants that run a paper travel card with the valve.
- Findable by valve, not by date. When a customer calls about valve serial 4471, the record is retrieved by searching for it — history query by valve identity, not scrolling through a month of runs.
- Baselined at the FAT. The FAT exports are kept as the machine's acceptance baseline; a year later, a wrong-looking reading is compared against the baseline to separate calibration drift from genuine valve rejects. That is the argument set out in the FAT checklist.
- Named so the archive sorts itself. Machine serial, valve serial, date — a file-naming habit costs nothing at export and is the difference between an archive and a folder of strangers.
What inspectors actually ask for
Third-party surveyors and incoming-inspection departments converge on the same three asks, in this order:
- Show me the curve of this valve. Matched by serial, produced during the hold — not a template report filled in afterwards.
- Show me the criterion it was judged against. Test pressure, hold time, acceptance factor or rate class — stated on the record.
- Show me the machine's own baseline. Occasionally, on a longer audit: the FAT record of the bench itself, to confirm the instrument was in order when the valve was tested.
A bench that stores curves, exports them and searches by valve identity answers all three without a meeting. The pressures it must reach for each class are in the hydrostatic test pressure chart.
Frequently asked questions
Why is the time-pressure curve more convincing than a single pressure reading?
A single end-of-hold reading only says where the pressure was at one instant. The curve shows the whole hold: whether the pressure was reached smoothly, whether it stayed flat, and if it fell, how fast and from where. Two valves can both show the same end pressure — one held flat, the other dropped from higher and only just cleared the margin. Only the curve tells those tests apart.
What should a complete valve test record contain?
Eight fields as a minimum: valve identity (type, DN, class, serial or batch), machine model and serial, test medium, test type (shell / seat / back-seat), set pressure and hold time, the time–pressure curve, the verdict against a stated criterion, and operator and date.
What does an instant pressure drop mean, versus a slow decay?
An instant step drop usually points at the fixture — a plate that slipped or a seal that let go — so re-clamp and run again before condemning the valve. A slow, continuous decay through the hold points at real leakage through the seat or a body defect, and that is a valve question, not a machine question.
How long should test records be kept, and where?
At least as long as the warranty on the valves you ship, and longer if your customers' industries demand it — the record is your defence if a field failure is traced back to the works test. Keep one copy outside the machine: every Changxun bench stores runs with curves in the controller and exports over USB, so the archive lives on your network.
Can the records be exported for our customers' incoming inspection?
Yes. Runs are stored with their curves and exported over USB as files that open on any office computer, so a customer's incoming-inspection department receives the curve of the exact valve they received — matched by serial number, not a template report.
Should the acceptance criterion be written on the test record?
Yes. The criterion — test pressure, hold time, acceptance factor or leakage rate class — is agreed before the run and stated on the record; for a PN-rated valve, the shell test at 1.5 × PN and the seat test at 1.1 × PN with the hold per the applicable standard. A verdict without a stated criterion is an opinion.
Benches that produce records like these
DFS2-600 · Dual station
- DN50 – DN300 / DN350 – DN600
- 0 – 7 MPa water
- Curve per run, USB export, history query
SYJ-600SQ · Water + air
- DN50 – DN600
- 40 MPa water / 4 MPa air
- Medium per record — water and air results never mixed
ZZC2-200 · Automatic line
- DN50 – DN200 gate valves
- Strength / lower seal / upper seal sequence
- Printed test records at the station
SYJ-150 · Compact
- DN15 – DN150
- 0 – 40 MPa, 30 t
- Small-bore high-pressure records for repair shops
For the full selection path — pressure, leakage class, layout, fixture, acceptance and now records — start at the valve pressure test bench buyer guide, then the test pressure chart, the leakage rate classes, the clamping methods reference and the FAT checklist.
Ask for a sample test record with its curve. Send your valve types, DN range and test pressure — the proposal includes the model, the fixture and the record format, within 12 working hours.