Surrogate modeling of additive manufacturing of metamaterials
Loading...
Date
2025-02-27
Authors
Κρυπαράκου, Θεοδώρα
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Lattice structures belong to the metamaterial family which are artificially made materials with mechanical properties that are not found in natural materials. They have gained popularity in recent years due to their attractive mechanical properties and the advancements in AM technologies. Changing the geometry and material parameters of a lattice RVE changes the effective mechanical properties of it. This possibility to alter their mechanical properties allows the mechanic to tailor them for case-specific designs. This study focuses specifically on the Body Centered Cube (BCC) lattice element. This is is a very interesting type of unit cell geometry as it combines a very easy to manufacture geometry with structural efficiency.
Finding the best design for a case-specific problem is accomplished through an optimization problem, which is a mathematical framework that allows for easy location of the optimal design by scanning the design parameter range. This thesis aims to solve a multi-objective optimization problem focuses on the BCC lattice RVE. The goal is to simultaneously minimize the weight of the RVE (important for lightweight applications) and the deflection of the struts of the BCC RVE under compressive load (important for maintaining structural integrity). The optimization problem is solved using a genetic algorithm (specifically in this thesis MOGA II).
Three different models are created for the purpose of solving the optimization problem, each with its own advantages and limitations. The first model uses analytical expressions, providing a fast and intuitive way to approximate the RVE’s behavior under load. The second model utilizes Finite Element Analysis (FEA), significantly improving accuracy but at the cost of increased computational time. Given the high computational demands of FEA-based optimization, a surrogate model (metamodel) trained on FEA results is introduced. This metamodel is used to perform a virtual optimization, replacing direct FEA evaluations with AI-based predictions, drastically reducing computational cost and runtime.
The results highlight the strengths and limitations of each approach, showcasing the trade offs between accuracy, computational efficiency, and generalization capabilities. This exploration of modeling trade-offs provides valuable insights into how different approaches can be used for lattice structure optimization in future work.
Description
Keywords
Lattice, Representative volume element (RVE), Surrogate models, Metamaterials