Exploiting the use of deep learning techniques to identify phase separation in self-assembled microstructures with localized graphene domains in epoxy blends

In this study for the first time, convolutional neural networks (CNN) are applied to identify phase-separated microstructures in a novel nano-modified polymer composite. The input data consist of a mixed labelled dataset of homogeneous microstructure and phase-separated microstructure images collected using an optical microscope. The problem is modelled as a binary classification and the initial model accuracy is 65.4% when the limited dataset is expanded using a data augmentation technique. When the dataset is increased by generating more ground truth images from experiment, the model is tested against a real dataset consisting of microstructure images of mixed structures and could classify them with an accuracy of 80.1%. Using the trained model, the microstructure is classified in a matter of minutes, saving hours of manual screening. This could be further useful in inverse design of phase-separated microstructure to predict the required composition for phase separation, saving several hours of manual work based solely on experimental methods. The problems in leveraging the ability of data-driven methods for polymer nanocomposites and challenges of data augmentation methods using a scarce dataset are discussed.