Fluid data consistency testing incorporation into the ThermoDataEngine

Transcription

1 Fluid data consistency testing incorporation into the ThermoEngine Vladimir Diky 2, John O'Connell 1 *, Jens Abildskov 3, Kenneth Kroenlein 2, Michael Frenkel 2 1 Professor Emeritus, University of Virginia, Nipomo, CA 2 National Institute of Standards and Technology, Boulder, CO 3 CAPEC-PROCESS, Technical University of Denmark, Lyngby, DK

2 Thermophysical Property Present in most separation models Quality of great concern Literature: Andrew R. Nelson, Jon E. Olson, Stanley I. Sandler, "Sensitivity of Distillatlon Process Design and Operation to VLE ", Ind. Eng. Chem. Process Des. Dev., 22(3), 548. Wallace B. Whiting, "Effects of Uncertainties in Thermodynamic and s on Process Calculations", J. Chem. Eng., 41, Paul M. Mathias, "Sensitivity of Process Design to Phase Equilibrium - A New Perturbation Method Based Upon the Margules Equation", J. Chem. Eng. 2014, 59, Enhanced rate(s) of publication creates new challenges

3 Thermophysical Property Amount doubles each decade ~30% papers contain errors in values, metadata, & uncertainties Errors propagate & cause damage in analyses & models validation essential in critical evaluation to: Identify questionable data Prevent erroneous use Increase value of Information about accuracy & reliability Knowledge to enforce consistency among properties Wisdom to reliably support applications NIST Thermo Engine (TDE) implements methods

4 Liquid Volumetric No rigorous consistency test (e.g., Gibbs-Duhem Equation used for partial molar properties) Validation currently involves finding: Outliers via Examining trends (e.g., density vs. pressure) Comparing independent measurements Systematic issues via models Compressed liquids: e.g.,tait model or multiparameter EOS Unsatisfactory if no theoretical basis or too many parameters Present work describes successful use of 3-parameter Corresponding States for Densities > 1.5 critical Temperatures < T critical

5 Correlation of Reduced Bulk Modulus Reduced Bulk Modulus, B, related to integral, C, of molecular direct correlation function, c, (DCFI) P / RT 1 B 1 c( r;t, ) dr 1 C T, T RT x Corresponding States Correlation for C (Huang, 1987) P / RT 1 C T, T 3 2 i j C * aij / * T / T * i0 j 0 a ij universal, insensitive T/T*, isotherm crossover at C* C*, *, T * fitted to over 300 substances & mixtures

6 Pressure-Density Pressure-density relation from integrating B P P0 x 1,, ix j Cij T x d RT 0 i j Given T, P, x, find from standard state of P 0, 0 for pure or mixed liquid at low P 0 Most reliable current model Benzene Mass Density, kg-m 3

7 Background. Xe with REFPROP Xe data: - Streett W. et al. (J. Chem. Thermodynamics, 1973): Greatest deviations from REFPROP predictions - Similar deviation from REFPROP EOS for high-pressure supercritical data (Michels et al., Physica (Amsterdam), 1954, 20: 99) Lemmon E.W., Jacobson R.T., J. Phys. Chem. Ref., Vol. 34(1) :з one or the other is erroneous Streett W. et al. Michels : Impeccable : World standard. EOS (Huang/O Connell, 1987): - Streett et al. > REFPROP Sending reliable data, but wanting to see if we picked up that REFPROP is not so good, though smooth Equation %-deviation (Street, 1973) %-deviation (Refprop) CS Discrepancy showed up as we might hope! REFPROP is easier to over-train!

8 Examples I. Accurate Xenon Benzene Density, kg-m Density, kg-m Deviation from correlation, kg-m ~0.25%

9 Examples II. Erroneous 990 Cyclohexanone No Negative Curvature for as Required 970 Density, kg-m

10 Examples IIIa. Publishing Typos Published Methanol/LiCl Repaired Density, kg-m Mass Density, kg-m Title in density Numbers in specific volume

11 Examples IIIb. Publishing Typos Toluene/SF Omit Misentry? Density, kg-m Regression Better Affects Regression Temperature, K

12 Examples IV. Multiphase Toluene/SF Multiphase Typo Mass Density, kg-m 3

13 Examples V. Inadequacy 1-Hexene 1200 Tait CSP 1000 P = MPa 800 P = MPa 800 Errors at High P Density, kg-m P = 0.1 MPa 600 P = 0.1 MPa Temperature, K Temperature, K

14 Summary Criteria o Low adequacy function, Q = N -1 Σw i (p calc -p exp ) i 2 o P(ρ) must be monotonically increasing o Abnormal parameter values (T i * T ci ; V i * V ci /3; C i * < 0) Detections o High quality work compromised by errors in manuscript preparation o Single phase data mixed with multiphase data o Adequacy to represent data of sufficient quality

15 Conclusions Liquid density data validation established Based on CSP formulation of reduced bulk modulus Allows uncovering of variety of errors Random & systematic measurement errors Textural & typographical reporting errors Implemented in NIST TDE software (trc.nist.gov/tde.html) References: Original CSP model Huang & O Connell Fluid Phase Equil. 1987, 37, Applied to ionic liquids - Abildskov, et al. Fluid Phase Equil. 2010, 295, TDE implementation - Diky, et al. J. Chem. Eng. ASAP 2015, DOI: /acs.jced.5b00477