Thermodynamic modeling and solubility assessment of oxycodone hydrochloride in supercritical CO2: Semi-empirical, EoSs models and machine learning algorithms

Abstract

In this study, the solubility of oxycodone hydrochloride (OXH) in supercritical carbon dioxide (SC–CO2) was investigated at various conditions, temperature (308–338 K) and pressure (120–270 bar), for the first time. The solubility ranged from 0.007 to 0.109 g/L, corresponding to mole fractions ranging from 0.051 × 10−5 to 0.699 × 10−5. Three different model groups were used to analyze the experimental data. The first group comprised seven semi-empirical models, with 3–6 adjustable parameters. These models include Sparks, Sodeifian 1 and 2, Bian, Jouyban, Gordillo and Jafari-Nejad. The second group employed two state equations, namely the Peng-Robinson (PR) and Soave-Redlich-Kwong (SRK) with van der Waals mixing rule. The average absolute relative deviation percentage (AARD%) was 9.73 and 10.63 for PR and SRK, respectively. The third group utilized four machine learning algorithms including DNN, RF, MLP and DTs with the respective R2 values 0.992, 0.980, 0.964 and 0.961, respectively. All of the models exhibited satisfactory agreement with the experimental data. Finally, the enthalpies of vaporization (79.71 kJ/mol) and solvation (−19.25 kJ/mol) were calculated for the first time.

Keywords

• Supercritical carbon dioxide (SC–CO2)
• Solubility
• Oxycodone hydrochloride
• Sodeifian models
• Machine learning algorithms
• Equation of state