PCDmanager Cp Data Fitting

[nrgeng]

From a previous posting, I have checked the regression statistics in what follows.

Ideal gas heat capacity

• selected estimating method from available choices in <T Correlations> panel
  imported data
  fit imported data

Seems to have worked, but regression statics are unverified by me at this time.

Ideal gas heat capacity: Original curve upon opening <T Correlations> panel

Original Curve

Compound W cp (scrn) 1.1.jpg (206.61 KiB) Viewed 8843 times

Ideal gas heat capacity: Imported external data

Imported External Data

Compound W cp load (scrn) 1.1.jpg (218.8 KiB) Viewed 8843 times

Ideal gas heat capacity: Fitted imported data

Fitted Imported Data

Compound W cp fit04 (scrn) 1.1.jpg (218.09 KiB) Viewed 8843 times

Analyzing the vapor heat capacity data using a second-order polynomial, least squares, linear regression provided different results from the PCDmanager regression statistics from the fitted imported data. My regression statistics were:

• coeff A = 16378
  coeff B = 563.30
  coeff C = -0.031591
  std err A = 16196
  std err B = 70.241
  std err C = 0.072218
  R2 = 0.99993
  std err Y = 2988.9
  F value = 27690
  Deg of Freedom = 6
  SSD mean = 7.4210E+11
  SSD given Y = 5.3601E+07

What is happening here? Why the difference?

[hkooijman]

The difference can be explained that in PCDman a version of the http://en.wikipedia.org/wiki/Levenberg% ... _algorithm is being used. This method is usually good but sometime - when starting from bad initial guesses - it is not. Particularly the initial guess method needs to be improved. This will be done over time.

[nrgeng]

Data from <Table> pane in <T Correlations> panel
(Analyzed data has units of T (K) & J/kmol/K)

PCDmanager Analysis of data

• CpIG Eqn 3 Data (Riazi-Daubert 1980)
  T min (K) 255.400
  T max (K) 922.00
  A -42463.55
  B 780.1198
  C -0.324541

nrgeng Analysis of data

• CpIG Eqn 3 Data (Riazi-Daubert 1980)
  T min (K) 255.400
  T max (K) 922.00
  A -42463.3705
  B 780.1195
  C -0.324541

• R2 = 1.0000
  std err Y = 0.36531
  SDD given Y = 2.4022

Very good data fit

Conclusion: PCDmanager fit of CpIG Eqn 3 Data (Riazi-Daubert 1980) is very good. Whereas the fit of Imported external data is not good. Why not use the same method, which is very good, for Imported Data as used for Estimating Method Data?