Session: 01-05-01: Applications of Artificial Intelligence/Machine Learning for Aerospace Structures 1

Paper Number: 162187

162187 - Deep Learning Model for Architected Interpenetrating Phase Composites With Targeted Stiffness 

Interpenetrating phase composites (IPCs), also known as co-continuous composites, contain at least two continuous and interconnected phases with three-dimensional (3D) topologies. In comparison with traditional particle or fiber reinforced composites, each phase of an IPC can stand alone and continue to carry loads even after one phase is damaged or removed. IPCs have been shown to exhibit superior mechanical properties and are finding important applications [1,2].

To achieve targeted properties, efforts have been made to optimize microstructural designs of IPCs. In addition to the conventional simple-cubic, body-centered-cubic and face-centered-cubic microstructures, triply periodic minimal surfaces (TPMS), which were inspired by natural materials, have also been employed to create IPCs. TPMS have the periodic topologies in three principal directions and display zero mean curvature at every point on the surface, leading to the minimal surface energy, surface stress and surface tension. As a result, TPMS-based cellular structures usually experience low stress concentrations. Recently, significantly higher stiffness and strength have been obtained for TMPS-based IPCs than those of the conventional truss-based composites due to the excellent mechanical properties of TPMS. Moreover, these TMPS-based IPCs exhibit excellent energy absorption, fracture toughness and recoverability.

It has been a great challenge to characterize microstructures of TPMS-based IPCs. The traditional approaches based on theoretical analysis, numerical simulation and topology optimization require expert knowledge and tend to be time-consuming. Recently, data-driven models have been proposed to accelerate the design and analysis, where expert knowledge is not a prerequisite. In addition, deep learning models, including the convolutional neural network (CNN), generative adversarial network (GAN) and deep neural network (DNN), were employed to predict mechanical properties. Furthermore, deep learning models have been used to perform inverse designs, which need much less computational resources than traditional analysis and optimization methods. However, owing to the cubic symmetry of TPMS, existing TMPS-based IPCs designed using the original TPMS functions display only properties of cubic materials. Hence, topology-property mapping for IPCs needs to be systematically studied. Also, data-driven models should be further explored to inversely design TPMS-based IPCs with targeted mechanical properties.

In this study, a deep learning model is proposed to design new IPCs with various material symmetries. Hybrid TPMS are mathematically designed as the reinforcement phase to construct the IPCs. The IPCs display orthotropic, tetragonal and cubic material symmetries, whose stiffness tensors are determined using a numerical homogenization method. A robust dataset is generated to establish the topology-property mapping. Then, a tandem dual-network model, including a forward sub-model and an inverse sub-model, is developed to predict the stiffness tensor and design the topologies. In addition, the hyperparameters are optimized to improve the accuracy of the model. The dual-network model provides excellent prediction and design capabilities. Moreover, the inversely designed IPCs can fulfill the requirements of the targeted stiffness both in and beyond the original dataset, which include orthotropic, tetragonal and cubic material properties. The current study provides a feasible approach to the forward prediction of elastic properties and inverse design of topologies through deep learning.

References

[1] Ai, L. and Gao, X.-L., 2017. Micromechanical modeling of 3-D printable interpenetrating phase composites with tailorable effective elastic properties including negative Poisson’s ratio. Journal of Micromechanics and Molecular Physics 2, 1750015-1~21.

[2] Ai, L. and Gao, X.-L., 2018. Evaluation of effective elastic properties of 3-D printable interpenetrating phase composites using the meshfree radial point interpolation method. Mechanics of Advanced Materials and Structures 25, 1241-1251.

Presenting Author: Xin-Lin Gao Southern Methodist University

Presenting Author Biography: Dr. Xin-Lin Gao is currently a professor of mechanical engineering at Southern Methodist University. He is a fellow of ASME and a past chair of the aerospace division of ASME.