A data-driven approach to full-field nonlinear stress distribution and failure pattern prediction in composites using deep learning

Abstract

An image-based deep learning framework is developed to predict nonlinear stress distribution and failure pattern in microstructural representations of composite materials in this paper. The work is motivated by the complexity and computational cost of high-fidelity simulations of such materials. The proposed deep learning framework predicts the post-failure full-field stress distribution and the crack pattern in two-dimensional representations of the composites based on their microstructures. The deep learning framework contains two stacked fully-convolutional networks, namely, Generator 1 and Generator 2, trained sequentially. First, Generator 1 learns to translate the microstructural geometry to the full-field post-failure stress distribution. Then, Generator 2 learns to translate the output of Generator 1 to the failure pattern. A physics-informed loss function is also designed and incorporated to further improve the performance of the proposed framework and facilitate the validation process. The material of interest is selected to be a unidirectional carbon fiber-reinforced polymer composite. 4500 microstructural representations are synthetically generated and simulated using an efficient finite element framework to provide a sufficiently large data set for training and validating the deep learning framework. It is shown that the proposed deep learning approach can predict the composites’ post-failure full-field stress distribution and failure pattern, two of the most complex phenomena to simulate in computational solid mechanics, with an impressive accuracy of 90%.