Bring your own binary VLE data
Your measurements.
Inspect, test, fit.
Paste a spreadsheet table or load a CSV. Inspect inferred activities, compare applicable consistency diagnostics, and fit NRTL or Margules to the same observations.
Start with the data
Enter sources, select units, then paste or load a table. Pure endpoints remain visible but cannot specify an absent component's activity coefficient.
Even measured pure endpoints do not give infinite-dilution activities. Extrapolation is an assumption. It never creates experimental observations.
Fit an activity model
Otherwise dimensionless parameters are constant over the measured T range. This is a chosen approximation, not evidence of temperature independence. Fits minimize mean [(100 ΔP/P)² + (100 Δy)²]; no uncertainty weighting or confidence intervals.
No fit yet.
Imported observations Normalized to kelvin and kPa; original table retained in JSON
What these tests can establish Scope and interpretation
Area and Herington calculations use the same teaching implementations as Lab 03. Without an explicit endpoint assumption, only the measured interval is integrated and no full-area criterion is assessed. For isothermal data, pressure/excess-volume terms are neglected. Herington's temperature-range term is empirical and is not a measured HE correction.
The differential diagnostic uses three-point derivatives on the actual composition spacing. It is sensitive to noise. For isobaric data, a verdict is withheld unless independent HE is supplied at every usable point. Supplying HE does not resolve excess-volume or vapor-fugacity approximations. No HE is inferred from the fitted model to certify its training data.
Model reproduction reports mean absolute pressure error (%) and mean absolute vapor-composition error in percentage points. The illustrative thresholds are each less than 1. Fits at parameter bounds are flagged. A fit may be poor because of model choice, property errors or measurement errors; it does not identify the cause. Fitted models are not automatically validated for extrapolation, stable flash prediction, or real-fluid process design.
Replicates at the same x₁ must be reviewed outside this importer. No silent averaging, endpoint filling, row deletion, outlier removal or unit guessing is performed. Rows outside either Antoine range block import; correlations are never extrapolated. Reimport after any data, unit or source change.
Calculation source · Regenerate study figure · NIST area/Herington guidance · NIST modeling-test guidance.