Stack delivered, guarantee test ran at conditions a few degrees or a few bar off reference. This corrects it back, propagates measurement uncertainty, and gives you a verdict under your contract's comparison convention — PASS/FAIL, or PASS/FAIL/INCONCLUSIVE — not a single number both sides will argue about.
Stack model
From a fitted curve (Tool 2) or typical defaults. This model is only used to compute the correction — the measured value itself is never altered.
j₀ is only meaningful at the temperature it was measured/fit at — this tool Arrhenius-scales it to your test and reference temperatures below. If you fit this from the Curve Fitter, use the temperature you fit at, not the guarantee temperature.
Defaults are a representative PEM stack, not your stack. Paste your own fit from the Curve Fitter before relying on this for a real acceptance test.
Left blank, the combined uncertainty below only reflects your instrument accuracy — not how well-determined j₀/α/ASR actually were. A correction built on "representative PEM defaults" has real model uncertainty even if you never fill this in; sending parameters from the Curve Fitter fills it automatically.
Test point
Measured vs. guarantee conditions
Guarantee & measurement uncertainty
k=2 ≈ 95% confidence interval, standard metrology convention (GUM).
Comparison convention
This is a contractual policy choice, not a physics question — pick the one your test procedure actually specifies.
Supplier must prove compliance even in the worst case within uncertainty (V_corrected + U ≤ guarantee → PASS) — the most buyer-conservative reading, and the default here. Not the only valid contractual convention; see Method below.
Fill in the test point and click Correct & evaluate.
One test, one snapshot, one argument settled. HYDRA OS runs this correction continuously against live operating data, so degradation and reference-condition drift are visible months before your next contractual test.
A witnessed acceptance test that ran 6 °C hot and 0.3 bar high on the anode, against a 55.0 kWh/kg guarantee. These are this page's default values.
Inputs
Measured
53.8 kWh/kg at 1.5 A/cm²
Faradaic efficiency
99.8 %
Test conditions
66 °C, 30 bar cathode, 1.3 bar anode
Guarantee conditions
60 °C, 30 bar cathode, 1.0 bar anode
Guarantee value
55.0 kWh/kg
Meter uncertainty
1.0 %
Temperature / pressure
±1.0 K / ±0.3 bar
Coverage factor
k = 2
Corrected result
Measured cell voltage
2.0192 V
Temperature term
+20.9 mV
Cathode pressure term
0.0 mV
Anode pressure term
−1.9 mV
Corrected voltage
2.0383 V
Corrected result
54.31 kWh/kg
Expanded uncertainty
±1.10 kWh/kg (k = 2)
Uncertainty band
53.21 … 55.40 kWh/kg
Same stack, same test, same numbers — and the verdict depends entirely on which convention the contract specifies: PASS under point comparison, PASS under benefit-of-doubt-to-supplier, INCONCLUSIVE under benefit-of-doubt-to-buyer.
Running 6 °C above the guarantee condition flattered the raw reading by 20.9 mV, worth about half a kilowatt-hour per kilogram — enough on its own to turn a marginal fail into an apparent pass. But the more consequential number here is the ±1.10 kWh/kg uncertainty band, which straddles the 55.0 guarantee line. Under the buyer convention that is an honest INCONCLUSIVE: this test, with this instrumentation, cannot support a PASS or a FAIL claim. Tightening the flow meter from 1.0% to 0.5% is what closes that band — and that is a decision to make before the witnessed test, not after.
Method & sources
This tool corrects a measured value to reference (guarantee) conditions using the same Butler-Volmer cell-voltage model as the Curve Fitter — the correction is the model's predicted voltage difference between your test conditions and guarantee conditions, added to the measured value. The measured value itself is never edited or overridden.
The total correction is broken into temperature, cathode-pressure and anode-pressure components applied one at a time, in that stated order — the same convention used in ASME Performance Test Code correction curves for turbines and compressors. The order matters slightly for the split between components (though not for the total), which is exactly why the order is stated rather than left implicit.
Uncertainty propagation
Combined uncertainty follows the root-sum-square (RSS) method from the ISO/JCGM "Guide to the Expression of Uncertainty in Measurement" (GUM) — the standard reference for combining independent measurement uncertainties:
u_c² = (∂V/∂T · u_T)² + (∂V/∂p_cat · u_pcat)² + (∂V/∂p_an · u_pan)² + (u_meter · V_measured)²
+ (∂V/∂j₀ · u_j₀)² + (∂V/∂α · u_α)² + (∂V/∂ASR · u_ASR)² [model terms, only if supplied]
U = k · u_c [k=2 ≈ 95% coverage, standard GUM convention]
Partial derivatives are evaluated numerically from the same model used for the correction itself, at your measured (sensor terms) or reference (model terms) operating point — not approximated from a generic sensitivity table. The model-parameter terms only appear if you've filled in j₀/α/ASR uncertainty (auto-filled when you arrive from the Curve Fitter, which already computes them from the fit's own covariance matrix). Leave them blank and the combined uncertainty reflects instrument accuracy only — which understates the true uncertainty whenever the stack model itself is loosely constrained, most obviously when you're using the "representative PEM defaults" rather than your own fitted curve.
The verdict — three contractual conventions, not one physics answer
Comparing a corrected value plus uncertainty against a guarantee is a policy choice about who bears the risk of measurement uncertainty, not something physics decides. This tool supports three:
Point comparison (ASME PTC style):
V_corrected ≤ guarantee → PASS, else FAIL. U is reported, not applied to the verdict.
Benefit of doubt to buyer (default):
V_corrected + U ≤ guarantee → PASS
V_corrected − U > guarantee → FAIL
otherwise → INCONCLUSIVE
Benefit of doubt to supplier:
V_corrected − U ≤ guarantee → PASS
V_corrected + U > guarantee → FAIL
otherwise → INCONCLUSIVE
ASME Performance Test Code practice for turbines and compressors is closer to point comparison — correct to reference conditions, compare directly, and negotiate the uncertainty band separately in the test agreement rather than folding it into an automatic verdict. The two "benefit of doubt" conventions are a common contractual shorthand when the test procedure wants a single automatic PASS/FAIL/INCONCLUSIVE output; which one your contract actually specifies is exactly the kind of detail worth confirming before a witnessed test, not after. INCONCLUSIVE (under either uncertainty-based convention) is a real, useful result — it means the band straddles the guarantee line, so neither a PASS nor a FAIL claim is honestly supportable from this test alone.
What this tool deliberately does not do
It does not know your actual instrument calibration — the meter/sensor uncertainties are exactly what you enter, nothing more.
It does not include model uncertainty unless you supply it — see "Uncertainty propagation" above.
It does not correct for degradation between commissioning and test date — if meaningful time has passed, factor that in separately (see the LCOH tool's degradation model).
It does not pick your contract's comparison convention for you — the three options above are real, different, and equally "correct" mathematically; only your test agreement says which one applies.
It is not a substitute for a witnessed, contractually-defined test procedure — it is the correction math that procedure should be running.