Scholarlink

<p>This study proposes a computationally efficient multi-objective optimization (MOO) framework for structural design, aiming to reduce weight and improve fatigue durability in complex dynamic systems. Conventional optimization approaches rely on high-fidelity dynamic simulations, resulting in excessive computational cost and limited scalability. To overcome these limitations, the proposed method employs a response surface model (RSM) formulated using co-gradient functions under a frequency-separation constraint (FSC), enabling accurate geometric approximation of dynamic-stress responses with minimal computational effort. The framework integrates co-gradient-guided design of experiments (DOE), neural-network-based surrogate modeling, and the NSGA-II algorithm for Pareto-based optimization. A sensitivity-ratio criterion is introduced to quantitatively determine when the co-gradient approximation can reliably replace full dynamic analysis, ensuring equivalence to the conventional formulation within the frequency-separated domain. Application to the electric power-unit structure of a heavy-duty battery-electric truck demonstrated a 17% weight reduction (11.9&#xa0;kg) and a 55% increase in fatigue life (from 2.7 AET to 4.2 AET), while reducing optimization time by 90% (from 29 days to 2.5 days). These results confirm that the proposed framework efficiently balances competing objectives while maintaining accuracy and scalability, offering a practical high-performance solution for real-world engineering design.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Multi-Objective Optimization Approach for Lightweighting and Durability Improvement of Vehicle Structure with Co-Gradient Functions

  • 期刊论文

  • 发表日期：

    2026年5月25日

查看全文
  • Hyosig Kim

摘要

This study proposes a computationally efficient multi-objective optimization (MOO) framework for structural design, aiming to reduce weight and improve fatigue durability in complex dynamic systems. Conventional optimization approaches rely on high-fidelity dynamic simulations, resulting in excessive computational cost and limited scalability. To overcome these limitations, the proposed method employs a response surface model (RSM) formulated using co-gradient functions under a frequency-separation constraint (FSC), enabling accurate geometric approximation of dynamic-stress responses with minimal computational effort. The framework integrates co-gradient-guided design of experiments (DOE), neural-network-based surrogate modeling, and the NSGA-II algorithm for Pareto-based optimization. A sensitivity-ratio criterion is introduced to quantitatively determine when the co-gradient approximation can reliably replace full dynamic analysis, ensuring equivalence to the conventional formulation within the frequency-separated domain. Application to the electric power-unit structure of a heavy-duty battery-electric truck demonstrated a 17% weight reduction (11.9 kg) and a 55% increase in fatigue life (from 2.7 AET to 4.2 AET), while reducing optimization time by 90% (from 29 days to 2.5 days). These results confirm that the proposed framework efficiently balances competing objectives while maintaining accuracy and scalability, offering a practical high-performance solution for real-world engineering design.