Evolutionary de-homogenization using a generative model for optimizing solid-porous infill structures considering the stress concentration issue

Shuzhi Xu, Hiroki Kawabe, Kentaro Yaji* · 2024

Read Evolutionary de-homogenization using a generative model for optimizing solid-porous infill structures considering the stress concentration issue by Shuzhi Xu, Hiroki Kawabe, Kentaro Yaji* (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-19154.txt.

Introduction

Porous infill inside a part saves weight, but designing it well is treacherous: coarse pixel models misjudge both the final geometry and the stress peaks where failure starts. Shuzhi Xu, Hiroki Kawabe, and Kentaro Yaji propose evolutionary de-homogenization. Their framework keeps two faithful versions of the design in step: a low-fidelity density model for fast searching, and a high-fidelity model with exact geometry for true stress analysis.

A mapping links the two levels one to one, and a deep generative model joined with a multi-objective evolutionary algorithm steers the cheap variables toward better real structures. Test cases show the method delivering hybrid solid-porous designs with accurate boundaries and controlled stress. The paper appears in 2024.

Catalog record

CREATORShuzhi Xu, Hiroki Kawabe, Kentaro Yaji*
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-19154

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Evolutionary de-homogenization using a generative model for optimizing solid-porous infill structures considering the stress concentration issue Shuzhi Xu, Hiroki Kawabe, Kentaro Yaji* Abstract The design of porous infill structures presents significant challenges due to their complex geometric configurations, such as the accurate representation of geometric boundaries and the control of localized maximum stress. In current mainstream design methods, such as topology optimization, the analysis is often performed using pixel or voxel-based element approximations. These approximations, constrained by the optimization framework, result in substantial geometric discrepancies between the analysis model and the final physical model. Such discrepancies can severely impact structural performance, particularly for localized properties like stress response, where accurate geometry is critical to mitigating stress concentration. To address these challenges, we propose evolutionary de-homogenization , which is a design framework based on the integration of de-homogenization and data-driven multifidelity optimization. This framework facilitates the hybrid solid-porous infill design by bridging the gap between low-fidelity analysis and high-fidelity physical realizations, ensuring both geometric accuracy and enhanced structural performance. The low-fidelity level utilizes commonly used density control variables, while the high-fidelity level involves stress analysis based on structures with precise geometric representations. By employing a de-homogenization-based mapping method, a side-by-side correspondence between low-fidelity and high-fidelity results is established. The low-fidelity control variables are iteratively adjusted to optimize the high-fidelity results by integrating deep generative model with multi-objective evolutionary algorithm. Finally, numerical experiments demonstrate the effectiveness of the proposed method. Journal: Elsevier 1 Introduction Porous structures, due to their superior mechanical properties such as ultra-lightweight, high stiffness, and energy absorption capabilities, have garnered widespread attention in practical applications [ 1 ] [ 2 ] . Concurrently, advancements in Additive Manufacturing (AM) have significantly accelerated the development of porous structures by reducing material usage and shortening production time [ 3 ] [ 4 ] [ 5 ] [ 6 ] . In recent years, topology optimization [ 7 ] [ 8 ] [ 9 ] , a powerful computational tool for determining the optimal distribution of materials within a design domain, has been widely applied in the efficient design of porous structures. Despite these advances, purely porous infill designs often exhibit inferior mechanical performance compared to solid designs under the same conditions. Therefore, when emphasizing the advantages of porous infill designs, the focus often shifts toward their scalability for multifunctional applications or their potential for further weight reduction in fixed geometric configurations. To balance weight reduction and mechanical performance, the hybrid solid-porous design approach has emerged. This method improves both local and overall performance by introducing solid components into critical regions of the lattice structure. Researchers have implemented and validated these designs using topology optimization techniques, and results show that their performance surpasses that of purely porous infill designs [ 10 ] [ 11 ] [ 12 ] [ 13 ] . Currently, there are several methods for generating porous infill structures. A classical method is multi-scale topology optimization, which is closely integrated with the traditional topology optimization framework [ 14 ] [ 15 ] . In multi-scale optimization, the analysis model is divided into macro and micro scales. Each element in the macro model represents a microstructure, which is modeled as a heterogeneous material unit using homogenization theory [ 16 ] [ 17 ] [ 18 ] . Many studies have explored how to optimize structures on both scales simultaneously [ 19 ] [ 20 ] . However, this method has encountered bottlenecks due to its high computational cost, leading to a slowdown in research. To address this issue, various compromise methods have emerged, such as parameterized unit cells with single or multiple parameters [ 21 ] [ 22 ] [ 23 ] and multi-domain parallel optimization [ 24 ] [ 25 ] [ 26 ] [ 27 ] . Recently, with the development of deep learning models, multi-scale optimization has regained attention under the concept of free material design [ 28 ] [ 29 ] [ 30 ] [ 31 ] . Compared to traditional solid structure optimization only in macroscale, multi-scale optimization provides greater design freedom, allowing the approach to reach theoretical limits and enabling non-conventional designs. However, this increased design freedom often results in overly complex geometries that are difficult to manufacture using mainstream methods. For instance, structures generated through multi-scale design often contain many microstructures with extremely small geometric features, making precise manufacturing more complex [ 15 ] . Additionally, due to their inconsistent multi-scale nature, these complex structures are difficult to interpret as coherent and manufacturable designs [ 15 ] . A simple solution is to scale each microstructure to the size of bilinear elements and apply appropriate orientation. The process of generating coherent and physically feasible designs through homogenization-based optimization is called de-homogenization. As a representative de-homogenization technique, Pantz and Trabelsi [ 32 ] proposed a post-processig method that provides an implicit geometric description for obtaining well-connected single-scale designs from spatially varying multi-scale designs. Recently, this method has been simplified and improved [ 33 ] [ 34 ] , and it has been applied to various problems [ 35 ] [ 36 ] [ 37 ] . Another method involves imposing local volume constraints to control local material distribution. Wu et al. [ 38 ] introduced upper bounds on the proximity of each design element, with constraints on localized material volume fraction, and aggregated all local per-voxel constraints using the P-norm. This idea has been extended to implement maximum length scale constraints in topology optimization, achieving bone-like porous structures. The application of localized material volume constraints in generating porous infill designs has since been widely discussed and applied to various scenarios, such as self-supporting infill structures [ 39 ] , multi-material structures [ 40 ] , stress-based designs [ 41 ] , and buckling designs [ 42 ] . To fully explore the weight reduction potential of sandwich structures while ensuring mechanical performance, shell-infill features have been incorporated into the topology design model. To the best of the authors’ knowledge, Clausen et al. [ 43 ] were the first to combine the density method with a topology design method for sandwich structures, utilizing a two-step density filter to describe the shell and infill domains. Subsequently, Luo et al. [ 44 ] proposed an erosion-based interface identification method to distinguish between the shell and infill layers in sandwich structures. Yoon and Yi [ 45 ] later developed a new density filter, which generates the sandwich structure by multiplying the modified density design variables with the original design variables. Besides the density method, such designs can also be realized through other approaches [ 46 ] [ 47 ] and can be combined with the aforementioned porous design methods to achieve more complex designs [ 48 ] [ 49 ] [ 50 ] [ 51 ] [ 52 ] . Significant challenges remain with current infill design methods using mainstream approaches. While traditional multi-scale optimization provides the greatest design freedom and the most rigorous theoretical foundation, the full-scale structures are often overly complex, making geometric representation difficult, especially in freeform design domains. In contrast, methods based on local volume constraints can accurately generate porous structures in macroscale. However, as a full-scale design method, they also require higher computational cost

Read the complete text