Open Access Open Access  Restricted Access Subscription Access

Machine Learning Based Structural and Failure Characteristics of Carbon Nanotube Junction Configurations

VINU UNNIKRISHNAN, VIKAS VARSHNEY, LANDON GABER, JAMES CASTLE, AJIT ROY

Abstract


Carbon nanotube based multi-terminal junction configurations are of great interest because of their potential aerospace and electronics applications. The performance of nanodevices based on these carbon nanotube junctions can be affected by the structural properties of the topologically accurate junction morphology like mechanical strength, thermal and electrical conductivities. It is therefore necessary to be able to investigate and characterize these properties accurately and quickly such that optimum structural configurations can be intelligently configured for the desired function. It is in this scenario that traditional atomistic simulations encounter severe limitations in terms of computational time required and hence the use of artificial intelligence techniques becomes necessary. In this work, we present a Machine Learning based characterization of structural and failure properties of carbon nanotube junction configurations. Structural and failure properties of a large set of 3D nanotube (CNT) junction configurations are first studied using molecular dynamics (MD) simulations. The stress-strain data until failure for each of the carbon nanotube junction configuration was estimated and various structural properties were used in the generation of the dataset. This dataset was used in a gradient boosting-based machine learning algorithm for the characterization of mechanical and failure properties of the nanotube junctions. Gradient boosting algorithms are prediction-based models that consists of sequential ensemble learning techniques in which the overall performance of the model is enhanced over multiple iterations. The coefficient of determination (R2) value is used to evaluate the robustness of the developed models for structural and failure parameters identified earlier. The machine learning techniques can be effective in predicting the structural properties of nanotube junction configurations with greater confidence.


DOI
10.12783/asc38/36717

Full Text:

PDF

Refbacks

  • There are currently no refbacks.