CNN-Informed Genetic Algorithm for Optimizing Mechanical Performance of Carbon Nanotube Microscale Bundles

In this study, a genetic algorithm is used to find carbon nanotube bundle microstructures whose bulk elastic-orthotropic properties optimally match desired bulk mechanical properties. Previous molecular dynamics simulations provide material properties for the carbon nanotube bundles. Three-dimensional finite-element micromechanics analyses of multi-bundle microstructures are performed with periodic boundary conditions to yield bulk elastic-orthotropic properties. Many configurations of carbon nanotube bundle microstructures are provided to a convolutional neural network for training. The convolutional neural network is incorporated into a genetic algorithm that searches the bundle microstructure design space for configurations that optimally achieve target bulk properties. The optimized configurations are re-analyzed via finite element simulations to assess the approach.