Logo image
Multiscale structural optimization for applications in thermal stability and actuation
Thesis   Open access

Multiscale structural optimization for applications in thermal stability and actuation

Isabella Snyder
Master of Science (M.S.), Drexel University
Jun 2024
DOI:
https://doi.org/10.17918/00010562
pdf
Snyder_Isabella_20244.43 MBDownloadView

Abstract

Actuators--Design and construction Neural networks (Computer science) Structural optimization--Industrial applications Thermal stresses
This thesis outlines a novel multiscale optimization framework for designing multifunctional structures that must withstand mechanical and thermal loads. The objective is to exploit the unique potential of spatially varying microstructures for enhanced thermal stability and actuation capabilities. The methodology hinges on a three-phase material design within the microstructure, composed of materials with high and low coefficients of thermal expansion (CTE) and void material, to achieve a spectrum of CTE from negative to positive. By optimizing the layout of these microstructures within a macrostructure, it is possible to induce desired thermomechanical behaviors, accommodating extreme conditions and precise deformations. To address the computational challenges inherent in designing complex multiscale structures, the research utilizes a deep neural network (DNN) surrogate model for numerical homogenization, significantly reducing the computational cost. The surrogate model predicts effective material properties, which are confirmed against traditional finite element analysis. The structure is optimized using an objective function and a constraint function. This paper analyzes how the use of compliance as the objective function and displacement as a component of the constraint function affects the results. Presented examples validate the optimization approach for thermal stability, where the target displacement is zero, and actuation, where the target displacement is non-zero.

Metrics

42 File views/ downloads
21 Record Views

Details

Logo image