Skip to content

Latest commit

 

History

History
 
 

README.md

Adjusting PC-SAFT parameters to experimental data

Notebooks

  • Adjust PC-SAFT parameters $m$, $\sigma$ and $\epsilon_k$ for a pure substance. 🠒 Notebook
  • Adjust binary PC-SAFT parameter $k_{ij}$ to VLE data. 🠒 Notebook
  • Adjust entropy scaling correlation parameters for the viscosity of a pure substance using PC-SAFT. 🠒 Notebook

Python scripts

Self-contained scripts fit the same hexane data (vapor pressure + liquid density) with the same initial parameters and Huber loss ($\delta = 0.05$), so their results and performance can be compared directly.

Script Solver Jacobian
fit_hexane_ad.py Levenberg-Marquardt via PureRegressor exact, via automatic differentiation
fit_hexane_scipy.py scipy.optimize.least_squares finite differences
fit_hexane_ad_scipy.py scipy.optimize.least_squares exact, via PureRegressor.predict

fit_hexane_ad_scipy.py uses PureRegressor as a compute engine only — data loading and AD-based gradient evaluation — while delegating the optimisation loop to scipy.