Modeling the thermophysical properties of alumina nanoparticles enhanced ionic liquids (NEILs) using advanced intelligent techniques

Abstract

Ionic liquids (ILs) are promising alternatives to conventional heat transfer fluids (HTFs) in thermal energy systems. This paper uses advanced machine learning (ML) approaches, specifically Cascaded Forward Neural Networks (CFNN) and Generalized Regression Neural Networks (GRNN), to predict the thermophysical properties of Alumina (Al₂O₃) nanoparticles in a binary mixture of water and the ionic liquid [C₂mim][CH₃SO₃]. Various optimization methods, including Bayesian Regularization (BR), Scaled Conjugate Gradient (SCG), and Levenberg-Marquardt (LM), were applied to enhance CFNN model performance. Alumina mass concentration and temperature were used as input parameters to predict specific heat capacity, thermal conductivity, and density, whereas shear rate and Alumina mass fraction were used for viscosity prediction. Results demonstrated that the CFNN model optimized with the LM algorithm closely matched experimental data, achieving average absolute percentage relative errors (AAPRE) of 0.2519 %, 0.2910 %, 0.0088 %, and 0.5937 % for specific heat capacity, thermal conductivity, density, and viscosity, respectively. Sensitivity analysis showed Alumina concentration strongly affected viscosity, density, and conductivity (r = 0.26, 0.92, 0.91), while temperature most influenced heat capacity (r = 0.74). Trend analysis showed that the CFNN-LM model captured the actual trends in the thermophysical properties of nanoparticle-enhanced ionic liquids (NEILs), and the leverage method validated the data, confirming its authenticity.