A Three-Pronged Computational Approach for Evaluating Density Based Semi Empirical Equations of Supercritical Extraction Process and Data

Software programs for parameter estimation, phase visualization and predictive modeling of supercritical extraction process and data using algorithms is presented in this work. A contextually appropriate, iterative, ordinary least squares estimation and selection method is developed for estimating model coefficients of density based semi empirical model equations associated with this process and data. Visualization of the phase behaviors projected by the specific density based semiempirical model equation(s) is also performed iteratively by plotting three-dimensional surfaces involving the state variables and solute solubility mole fraction. Predictive modeling of input empirical data has been implemented using three supervised machine learning algorithms (Multilayer perceptron, K-nearest neighbors and Support vector regression). Hyperparameter optimization of the machine learning algorithms is performed prior to prediction. Detailed analysis of the prediction is conducted by using standard scoring metrics and descriptive charts. Theoretical inference and discrepancies regarding the predicted window of maximum/optimal solubility, modeling efficiency, vapor liquid equilibrium and phase behaviors projected by the model equations have been elucidated from the program outputs. In summary, these programs are unique, accurate, reliable and simple computational tools for evaluating/designing density based semiempirical equation(s) of supercritical extraction process and associated data.