Thermophysical properties prediction of carbon-based nano-enhanced phase change material's using various machine learning methods

Background: In this modeling project, employing various machine learning methods, the thermophysical properties of phase change material (PCM) containing three nanoparticles were predicted. PCM with Multi-Walled Carbon Nanotube (MWCNT), Graphene Nanoplatelets (GNP), and nano-graphite (NG) were considered as nano-enhanced PCM, and their experimental data were extracted from the literature. Method: The data consisted of thermal conductivity and latent heat of 100 thermal cycles measured for 1, 2, and 3 wt.% of nanoparticles in the myristic acid (MA) as PCM. This research uses supervised algorithms such as k-nearest neighbors (KNN), Automatic relevance determination (ARD), and least absolute shrinkage and selection operator (LASSO) for regression, creating formulas and making models. Findings: This research using the ARD regression to find the relationship between solid-state and liquid-state for three materials (GNPs/MA, MWCNTs/MA, and NG/MA) with the R-Squared value of 0.999 for all materials. The MSE of the ARD algorithms for the materials respectively is 1 × 10⁻⁹, 9 × 10⁻¹⁰, and 6 × 10⁻¹⁰. Using the MA, which is the primary material, creates the polynomial regression for GNPs/MA, MWCNTs/MA, and NG/MA, and the R-Squared values are 0.981, 0.984, and 0.981 and the MSE, respectively is 0.000159351, 0.000016945, and 0.000022425. The KNN algorithm is used to make the model for this subject, and the R-Squared values are 0.985, 0.988, and 0.988. The LASSO regression is used to make the linear regression for the relationship between melting-state and freezing-state, and the R-Squared value is 0.999 for all materials. The MSE of the LASSO method for these materials of this part respectively is 0.000176203, 0.000000545, and 0.000005035.