Article: Comparing open-source optimisation algorithms for functionally graded material design: a thermoelastic case study Journal: International Journal of Computer Aided Engineering and Technology (IJCAET) 2024 Vol.19 No.1 pp.1 - 12 Abstract: In this study, we apply black-box optimisation (BBO) techniques using an open-source BBO framework, Optuna, to optimise the material composition of functionally graded materials (FGMs), specifically targeting residual thermal stress reduction in a uniformly cooled multi-layered FGM plate. We focus on three algorithms with an aim to compare their performance: the tree-structured Parzen estimator (TPE), the covariance matrix adaptation evolutionary strategy (CMA-ES), and the non-dominated sorting genetic algorithm II (NSGA-II). Our findings indicate that CMA-ES excels in optimisation quality, outperforming TPE and NSGA-II, despite TPE's rapid convergence. We also observe that accounting for interactions among design variables may not always be beneficial and can hinder the optimisation process. This study not only showcases the effectiveness of BBO in material science but also guides material designers in selecting suitable optimisation techniques for complex engineering challenges. Inderscience Publishers - linking academia, business and industry through research

Title: Comparing open-source optimisation algorithms for functionally graded material design: a thermoelastic case study

Authors: Ryoichi Chiba

Addresses: Department of Mechanical Engineering, Sanyo-Onoda City University, Sanyo-Onoda, 756-0884, Japan

Abstract: In this study, we apply black-box optimisation (BBO) techniques using an open-source BBO framework, Optuna, to optimise the material composition of functionally graded materials (FGMs), specifically targeting residual thermal stress reduction in a uniformly cooled multi-layered FGM plate. We focus on three algorithms with an aim to compare their performance: the tree-structured Parzen estimator (TPE), the covariance matrix adaptation evolutionary strategy (CMA-ES), and the non-dominated sorting genetic algorithm II (NSGA-II). Our findings indicate that CMA-ES excels in optimisation quality, outperforming TPE and NSGA-II, despite TPE's rapid convergence. We also observe that accounting for interactions among design variables may not always be beneficial and can hinder the optimisation process. This study not only showcases the effectiveness of BBO in material science but also guides material designers in selecting suitable optimisation techniques for complex engineering challenges.

Keywords: optimal design; functionally graded material; FGM; thermal stress; thermoelasticity; black-box optimisation; BBO.

DOI: 10.1504/IJCAET.2024.138652

International Journal of Computer Aided Engineering and Technology, 2024 Vol.19 No.1, pp.1 - 12

Received: 02 Sep 2023
Accepted: 07 Mar 2024
Published online: 22 May 2024 *

Title: Comparing open-source optimisation algorithms for functionally graded material design: a thermoelastic case study

Keep up-to-date