Hybrid Finite Element-Neural Network Method for Prediction of Bucky Paper Nanocomposite Elastic Properties

Porous assemblies of carbon nanotubes produced by filtration from dispersion is called bucky paper is likely to have broad application in future nanocomposites. Since bucky papers are porous materials when they are used in polymer nanocomposites the polymer chains place among carbon nanotubes. This matter affects bucky paper mechanical properties. In this study, a new method is presented for prediction of bucky paper mechanical properties considering polymer penetration effect. In the proposed technique, Finite Element (FE) method was used for 3D modeling of multi-wall carbon nanotubes (MWCNTs) embedded in polymer medium using periodic boundary condition. In this model, MWCNTs were considered as continuum shells with effective diameters and elastic modulus properties of atomistic models. The presented method is more accurate, simpler to implement as well as computationally cheaper than previously used Molecular Mechanics (MM) and Molecular Dynamics (MD) models. By carrying out FE simulations, effects of various polymer elastic moduli, MWCNTs diameters, gap distance and engagement percent on the bucky paper elastic modulus were discussed. Finally, a Multi-Layer-Perceptron (MLP) neural network is designed and trained, using the Levenberg-Marquardt algorithm, to approximate the results of the FE simulation. The trained network would be able to predict accurately, the elastic modulus of a certain bucky paper at a negligible computational cost

کلیدواژه ها:

Bucky Paper ، Finite Element ، Neural Networks ، Polymer Nanocomposites ، Elastic modulus