Explainable machine learning framework for CNT-PCM thermal conductivity prediction

Assessing the thermal conductivity of carbon nanotube-based composites is difficult due to the experimental uncertainties and complex thermal interactions. To address these challenges, this research presents an explainable artificial intelligence integrated artificial neural network-based deep learning framework (XAI-ANN-ML-F). For designing the proposed XAI-ANN-ML-F, experimental datasets (n=502) were inferred from the prevailing literature. The collected dataset represents single-walled carbon nanotubes (SWCNTs), multi-walled (MWCNTs) and functionalized multi-walled carbon nanotubes (FMWCNTs) to ensure a wide range of thermophysical characteristics and experimental conditions. The proposed approach includes three algorithms: Bayesian Regularization (BR), Levenberg-Marquardt (LM), and Scaled Conjugate Gradient (SCG). Among these, BR algorithm demonstrates the best predictive performance with a coefficient of determination (R2) of 0.96. Interpretability analysis found that the volume fraction, length and diameter of the CNTs are the major influential parameters governing keff/km. The proposed XAI-ANN-ML-F offers a reliable and effective tool for quick estimation and design optimization of CNT-based nanoparticles.