Thermodynamic Study of Propyl Propanoate and 1-Alkanols (C ₆ –C ₁₀ ): Application of Machine Learning for Prediction of Density of Mixtures

This study integrates experimental analysis and machine learning to characterize the thermophysical properties of ester-alkanol systems. Initially, densities and viscosities for five binary mixtures of propyl propanoate and 1-alkanols (C6–C10) were measured from 293.15 to 323.15 K. Positive excess molar volumes and negative viscosity deviations were obtained for all binary mixtures at all temperatures. Building upon these experimental insights, we applied three machine learning approaches named Support Vector Machine, Artificial Neural Network, and a Radial Basis Function Neural Network (RBFNN) to create a generalized predictive model for mixture density. The RBFNN architecture yielded the highest fidelity, achieving a Mean Absolute Percentage Error of 0.08% and a coefficient of determination (R2) of 0.9984. Notably, the final model demonstrates excellent transferability, accurately predicting densities for all five alkanol systems without system-specific adjustments.