Stochastic full-range multiscale modeling of thermal conductivity of Polymeric carbon nanotubes composites: A machine learning approach

Based on a stochastic full-range multiscale model, we propose a data-driven approach to predict the thermal conductivity of CNT reinforced polymeric nano-composites (PNCs). Uncertain input parameters at different scales are propagated from nano-to macro-scale within a bottom-up multi-scale framework. Atomistic models are employed at the nano-scale while continuum mechanics approaches are used at the micro-, meso-and macro-scale. Representative volume elements in the context of finite element modeling (RVE-FEM) are used to finally obtain the homogenized thermal conductivity. To connect the micro and mesoscale and simplify the computation, we take advantage of the equivalent fiber theory. The input parameters are selected by a top-down scanning method and subsequently are converted as uncertain inputs. The length of single-walled carbon nanotube (SWCNT), the chirality of SWCNT, the thermal conductivity of the fibers, the thermal conductivity of the matrix, the Kapitza resistance, aspect ratio, agglomeration index, dispersion index and volume fraction are assumed as random-parameters. The Regression-tree-based (Random Forest and Gradient Boosting Machine) and Neural networks-based (Artificial neural networks and Deep neural networks) approaches are exploited for computational efficiency, where Particle Swarm Optimization (PSO) and 10-fold Cross Validation (CV) are employed for hyper-parameter tuning. Our machine learning prediction results agree well with published experimental data, which can provide a versatile and efficient method to design new PNCs.

Automated summary

In this study, a data-driven approach based on a stochastic full-range multiscale model is proposed for predicting the thermal conductivity of CNT reinforced polymeric nano-composites (PNCs). Machine learning techniques are employed to efficiently design new PNCs based on the prediction results.