A vendor once quoted me a bench “rated to 600 psi, with a digital pressure readout to 0.1 bar resolution.” The acceptance criterion I had to enforce was a 0.014 MPa drop over five minutes. My transducer’s resolution was seven times larger than the signal I was trying to catch. Every “pass” was statistically indistinguishable from a fail.
That is the gap this article closes. It specifies the bench hardware and instrumentation that make a credible acceptance test possible — envelope, instrument accuracy classes, safety systems, fluid handling, and data acquisition. The methods themselves — hydrostatic, integrity air test, dynamic running — belong to the companion article on test methods for mechanical seals.
Pressure, Temperature, and Speed Envelope
The envelope is set by the seal categories the bench must qualify, not by the builder’s convenience.
Pressure and Temperature by Seal Category
API 682 Category 1 seals are specified from -40 to +260 °C (-40 to +500 °F) and up to 22 bar (315 psi). Category 2 and 3 push to -40 to +399 °C (-40 to +750 °F) and 42 bar (615 psi).
If your program covers Category 2/3, a 315-psi bench is out of spec the day it arrives.
Hydrostatic Headroom
Add headroom above the service envelope. The hydrostatic test required by API 682 4th Edition Section 10.3.3 holds the pressure vessel at 1.5× Maximum Allowable Working Pressure (MAWP) for 30 minutes without leaking.
A bench rated exactly to 42 bar cannot run a hydrostatic on a 42-bar seal — the pressurization circuit, vessel, and relief-valve setpoint must accommodate 63 bar (915 psi) sustained. Future ASME Section VIII revisions move toward 1.3×, but 1.5× is what the 4th edition demands.
Speed Envelope
Speed is the forgotten axis. A TEPEN research rig publishes a 120 to 1,500 rpm envelope in ten discrete steps with ±0.05 bar accuracy — tight numbers, but capped well below the 3,600 and 7,200 rpm speeds where API 682 qualification on OH2/OH3 pump configurations actually runs.
Specify maximum shaft speed as the highest RPM the target pump family reaches under relief-valve conditions, not the nameplate speed.
Instrumentation and Measurement Uncertainty
Your instrument accuracy must be smaller than your acceptance criterion. That one principle sets every other instrumentation number in this specification.
Accuracy Must Beat the Acceptance Criterion
API 682 and ISO 21049 allow a maximum 0.014 MPa (2 psi) pressure drop over 5 minutes from a 28 L reservoir at 0.17 MPa (25 psig) — 56.9 g/hr of air. A transducer with ±1% FS accuracy on a 0-35 bar range carries a ±0.35 bar uncertainty envelope, 25 times larger than the threshold.
You cannot declare a pass. Specify ±0.1% of reading or better, calibrated within six months, with a declared uncertainty budget on the certificate.
Transmitters, Not Switches
John Crane’s technical blog on the API 682 4th edition calls the shift from switches to transmitters with local indicators “the most obvious and significant” instrumentation change in the standard. A bench specified against third-edition assumptions — switch alarms, hand-calibrated gauges — will not qualify a seal against current acceptance criteria. Transmitters with a declared accuracy class are table stakes.
Reservoir Volume Changes the Signal
AESSEAL’s published testing pushes the point further. Their 1.5 L air-test reservoir produces a 22-psi pressure drop from a 0.008-inch orifice — 11× the allowable drop at the stricter volume. The same defect in a 26.5 L reservoir produces only a 2-psi drop and passes the standard test.
Whichever volume your procedure uses, the transducer must resolve the drop with meaningful headroom below the threshold, or the bench is sensitive to reservoir volume rather than to the seal.
Sensitivity vs Resolution vs Accuracy vs Precision
Four terms in procurement specs are routinely conflated. Sensitivity is the smallest input change the instrument responds to. Resolution is the finest increment displayed.
Accuracy is deviation from the true value, including gain error, offset, and noise. Precision is repeatability.
A transducer with 1 mV resolution but 15 mV sensitivity displays more decimal places than it can actually measure. Write the spec against all four.
Test Bench Safety Systems
A bench that can reach 63 bar during hydrostatic can kill someone when a seal ruptures.
Relief and Burst-Disk Sizing
Relief-valve setpoints must be sized for 1.5× MAWP overpressure duty, not working pressure. Burst disks belong downstream as a redundant element, sized to open before the vessel yields, and routed to a containment volume — not the operator side.
Containment and Shielding
The scale of a high-pressure failure deserves concrete numbers. AESSEAL’s testing of a 0.008-inch defect on an operating seal produced 1,500 g/h at 2 barg, 3,000 g/h at 7 barg, and 7,500 g/h at 40 barg — 270× to 1,300× the air-test allowable.
Now imagine the defect is a fractured face, the fluid is a hazardous substitute, and the bench is unvented. Shaft-containment sleeves sized to the full seal OD, face-shield glazing rated for working pressure, and hazard-classified vapor extraction are specification items, not nice-to-haves.
What the Integrity Test Does Not Validate
One subtle liability: the integrity air test validates cartridge assembly, not field performance. Gordon Buck’s work on the standard notes that the test allows a high leakage rate because it targets assembly defects — cut O-rings, scratched faces — not full pressure capability.
Bolts, gaskets, gland-plate contact surfaces, and the pump-chamber gasket pass the integrity test even when they would fail in service. Your safety envelope must contain the failure modes the test does not interrogate.
Test Fluid Handling
Dry-Face Discipline
Dry faces matter. A seal pre-oiled before an integrity air test reads near-zero leakage because the oil film closes the gap the test is designed to find.
Pre-oiling can pass a defective cartridge. Specify cleaning and de-oiling for incoming seals and any bench surface that contacts the faces before test. Before mounting, incoming mechanical seal inspection verifies face flatness and dimensional conformance — the bench cannot compensate for a seal that arrived out of tolerance.
Fluid Conditioning: Four Requirements
Fluid conditioning covers four requirements the generic “water loop” spec skips. Filtration to a declared particle size, typically 10 microns for water service and 5 microns for qualification. Temperature conditioning tighter than the acceptance criterion allows to drift.
Gas saturation or degassing depending on whether you simulate hot water or cold oil. Venting with provision for vapor capture if the substitute fluid is hazardous.
A bench that runs hydrocarbon substitutes for toxic or flammable services needs fluid-side venting classified to the hazard — not “routed to the shop roof.” Consult the ASME Pressure Vessel Code and a local AHJ before commissioning.
Test Bench Data Acquisition Requirements
Sample Rate per Channel
DAQ architecture depends on the acceptance criterion’s temporal scale. A 2-psi drop over 5 minutes means 6.67 mbar per minute — a 1 Hz sample rate resolves it easily. A rotordynamic pressure fluctuation at 3,600 rpm lives at 60 Hz and needs anti-alias filtering plus a sample rate of at least 2 kHz.
Specify sample rate per channel, not as a bench-wide number.
TEPEN’s published rig runs a 2 MHz primary DAQ at 16-bit for acoustic emission (100 kHz to 1 MHz), plus a 96 kHz secondary DAQ with 16 channels at 16-bit for pressure, speed, and temperature. That two-tier architecture — one fast channel for fault signatures, a slower multi-channel block for state variables — is a defensible baseline.
Multi-channel rates of 5 to 10 kHz across five or more strain or vibration channels are achievable on commodity telemetry and sufficient for rotor-dynamic diagnosis on most pump families.
16-Bit Resolution Minimum
Sixteen-bit resolution gives 65,536 counts per channel. On a 0-35 bar pressure channel, that is 0.5 mbar per count — below the 0.014 MPa (140 mbar) threshold by more than two orders of magnitude.
Twelve-bit resolution (4,096 counts) gives 8.5 mbar per count and is marginal. Specify 16-bit minimum for any channel tied to an acceptance criterion.
Data Retention for Audit
Data retention is the last mile. API 682 Annex I qualification testing — the API 682 mechanical seal standard provisions moved there from Annex F in the 4th edition — produces evidence packages that must survive audit for years.
Specify raw-data retention, open file format, channel metadata including calibration certificate reference, and time synchronization better than the shortest sample interval.
What Gets a Bench Rejected at Audit
The bench that fails a qualification audit usually has passing instruments on the wrong scale. A ±1% FS transducer on a 35-bar range cannot resolve a 0.014 MPa drop — the signal is inside the noise, and every “pass” is statistically indistinguishable from a fail.
The rule that keeps a specification defensible is single: instrument resolution must be smaller than the acceptance criterion, at every link in the chain — pressure, speed, temperature, sample rate, bit depth. Spec the envelope to the seal category at 1.5× MAWP, not to the builder’s convenience. Everything else follows from those two rules.
