Predicting thermophysical properties enhancement of metal-based phase change materials using various machine learning algorithms

Background: In this research, five machine learning methods are employed to create the formula and make the model for communication between the melting and solidification processes to find the latent heat and communication between the thermal conductivity and the thermal diffusivity. Paraffin wax was the PCM, while SiO2, ZnO, Fe2O3, and Al2O3 nanoparticles were utilized as additives to enhance the thermal conductivity by considering four concentrations between 2 and 8 wt.%.Method: The K-Nearest Neighbours (KNN), the decision tree, the stochastic gradient descent (SGD) regressor, the Support Vector Machines (SVM), and the Huber regression are the algorithms that are used in this study.Findings: Using the Huber regression method, the formula for the latent heat of the solidification-process is created by the melting-process (by latent heat and wt.) with the polynomial form by degree 2, and the R-Squared value is 0.998, and MSE is 0.971172, and by the KNN (K = 3) is made the model with the linear form. The R2 value is 0.999, and MSE is 0.41809, and for the latent heat of the melting-process by the solidification-process (by latent heat and wt.) with the same methods reached to best the R2 for creating the formula and making the model, respectively the R2 values are 0.996 and 0.999, and the MSEs are 1.482458 and 0.608658. Using the Huber regression method, reach the best result for creating the formula in the polynomial form by degree 2 for the thermal conductivity by the thermal diffusivity and wt. and create the formula with the R2 value of 0.998 and MSE is 0.000033, and make the model by the SVM with the R2 value of 0.999 in the linear form, and MSE is 0.000016. The same methods are used for thermal diffusivity by the thermal conductivity and wt.; the R2 values for the formula and the model, respectively, are 0.997 in polynomial form (degree = 2) and 0.999 in linear form. MSEs, respectively, are 0.000020 and 0.000006.