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

Paper Number: 121612

121612 - Lattice Structure Design Using Machine Learning and Homogenization Approach 

The homogenization approach is commonly applied to lattice structure design to simplify the modeling of complex geometric structures by using simple solid models. The homogenized material properties for the solid elements are obtained by using a Representative Volume Element (RVE). There are a number of common homogenization approaches in the literature, such as beam theory, Asymptotic Homogenization (AH), average volume method, etc., that are employed to determine effective material properties, including strength and stiffness constants. In our previous study [1], we used beam theory-based homogenization to study the static behavior of lattice plates with different commonly used lattice unit cells (Square, Triangular, Regular hexagonal, Mixed). Verification of the developed in-house Finite Element Method (FEM) code with existing literature [2] indicates that it can reduce computation time significantly while still maintaining excellent agreement with results obtained with a detailed finite element model in Abaqus. However, this approach can only be used for a limited number of representative shapes of lattice structures. Thus, it is necessary to develop a variety of analytical equations each time for any shape of lattice structure to optimize the structure by utilizing a homogenization approach. As a result, AH is now widely used to determine the mechanical properties of lattices. Since this approach can be applied to a wide range of shapes and does not depend on analytical equations, it outperforms the traditional beam theory-based homogenization approach. Despite its popularity in finite element analysis, this approach is expensive as it requires fine meshes to analyze smooth and complex lattice geometries. Therefore, it increases the complexity of the preprocessing and subsequent design analysis and optimization processes.

As part of our current study, we used this AH approach to determine the effective material properties of the lattice structures. Additionally, Image processing is employed instead of the traditional mesh generation process to construct the finite element model of the lattice RVE. Using this methodology, any image of a random-shaped RVE can be converted into a binary image. Consequently, this binary image is utilized to generate a binary matrix, which substitutes the traditional mesh generation process for calculating effective material properties. The effectiveness of this methodology has been verified by using results available in the literature [3] by comparing the effective material properties for different RVEs, which saves substantial preprocessing time in the computation of homogenized material properties. To enable a rapid determination of material properties using the image of the lattices, we are considering using the developed program to produce a data set that will be utilized in the training of a Neural Network (NN). Consequently, the NN will be capable of predicting the effective material properties of any lattice RVE shape from its image.

References:

1.            Mahdi, M.A. and W. Zhao, Stiffness Tailoring for Improved Bending and Buckling of Variable Angle Tow Composite Sandwich Panels, in AIAA SciTech 2024 Forum. Florida, USA.

2.            Kapania, R.K., et al., Static Analysis of Sandwich Panels with Square Honeycomb Core. AIAA Journal, 2008. 46(3): p. 627-634.

3.            Wang, C., et al., Concurrent topology optimization design of structures and non-uniform parameterized lattice microstructures. Structural and Multidisciplinary Optimization, 2018. 58(1): p. 35-50.

Presenting Author: Mohammed Abir Mahdi Oklahoma State University

Presenting Author Biography: Mohammed Abir Mahdi
Graduate Research and Teaching Assistant
Aerospace Structures and Materials Laboratory
Lab Website: https://www.sdolab.net/
325, Advanced Technology and Research Center (ATRC)
Mechanical and Aerospace Engineering (MAE)
Oklahoma State University
Stillwater, OK 74078