Accelerating Multi-Objective Optimization of Composite Structures Using Multi-Fidelity Surrogate Models and Curriculum Learning DOI Open Access
Bartosz Miller, Leonard Ziemiański

Materials, Год журнала: 2025, Номер 18(7), С. 1469 - 1469

Опубликована: Март 26, 2025

The optimization of multilayer composite structures requires balancing mechanical performance, economic efficiency, and computational feasibility. This study introduces an innovative approach that integrates Curriculum Learning (CL) with a multi-fidelity surrogate model to enhance efficiency in engineering design. A strategy is introduced generate training data efficiently, leveraging high-fidelity finite element for accurate simulations low-fidelity provide larger dataset at reduced cost. Unlike conventional modeling approaches, the proposed method applies CL iteratively refine model, enabling step-by-step learning complex structural patterns improving prediction accuracy. Genetic algorithms (GAs) are then applied optimize parameters while minimizing expense. integration allows reduction burden preserving accuracy, demonstrating practical applicability real-world design problems. effectiveness this methodology validated by evaluating Pareto front quality using selected performance indicators. Results demonstrate reduces achieving predictions, highlighting benefits integrating modeling, analysis, CL, GAs efficient structure optimization. work contributes advancement methodologies providing scalable framework applicable problems requiring high efficiency.

Язык: Английский

Accelerating Multi-Objective Optimization of Composite Structures Using Multi-Fidelity Surrogate Models and Curriculum Learning DOI Open Access
Bartosz Miller, Leonard Ziemiański

Materials, Год журнала: 2025, Номер 18(7), С. 1469 - 1469

Опубликована: Март 26, 2025

The optimization of multilayer composite structures requires balancing mechanical performance, economic efficiency, and computational feasibility. This study introduces an innovative approach that integrates Curriculum Learning (CL) with a multi-fidelity surrogate model to enhance efficiency in engineering design. A strategy is introduced generate training data efficiently, leveraging high-fidelity finite element for accurate simulations low-fidelity provide larger dataset at reduced cost. Unlike conventional modeling approaches, the proposed method applies CL iteratively refine model, enabling step-by-step learning complex structural patterns improving prediction accuracy. Genetic algorithms (GAs) are then applied optimize parameters while minimizing expense. integration allows reduction burden preserving accuracy, demonstrating practical applicability real-world design problems. effectiveness this methodology validated by evaluating Pareto front quality using selected performance indicators. Results demonstrate reduces achieving predictions, highlighting benefits integrating modeling, analysis, CL, GAs efficient structure optimization. work contributes advancement methodologies providing scalable framework applicable problems requiring high efficiency.

Язык: Английский

Процитировано

0