Pioneering machine learning techniques to estimate thermal conductivity of carbon-based phase change materials: A comprehensive modeling framework
Raouf Hassan, Alireza Baghban
Abstract
This study presents a comprehensive data-driven framework to accurately estimate the thermal conductivity of nano-enhanced phase change materials (NEPCMs) using machine learning. A dataset of 482 samples, incorporating various nanoparticle types, concentrations, PCM types, and operating temperatures, was curated and refined using the Monte Carlo Outlier Detection algorithm. Extensive machine learning algorithms were explored; however, CatBoost, XGBoost, ANN, Random Forest, and Gradient Boosting emerged as the most accurate models. Among them, CatBoost achieved the highest predictive performance with an R 2 of 0.979 and the lowest mean squared error (MSE) of 0.006 on the test set. SHAP analysis revealed that nanoparticle concentration was the most influential feature. These findings highlight the effectiveness of the proposed approach in capturing the complex interactions governing thermal conductivity in NEPCMs and provide valuable insights for material design and optimization.