Scholarlink

<p>To systematically enhance sustainability in automobile seat design and to holistically address the multi-dimensional requirements of the economic, environmental, and social dimensions under the Triple Bottom Line (TBL) framework, a data-driven design methodology integrating surrogate modeling with multi-objective evolutionary algorithms is proposed. First, user core needs are quantified and translated into 15 design features through SET (Social–Economic–Technical) factor analysis and the Fuzzy Analytic Hierarchy Process (FAHP). Subsequently, 10 key sustainability indicators are identified based on the TBL framework and categorized into economic, environmental, and social dimensions. Utilizing a dataset of 800 expert–student scoring records, global optimization of Random Forest (RF) hyperparameters is performed via Bayesian Optimization (BO), thereby constructing a high-accuracy surrogate model capable of predicting scores for any combination of design features across the identified sustainability indicators. The weighted composite scores for the three dimensions are adopted as optimization objectives, and the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed to explore Pareto-optimal solution sets for design feature configurations. Through high-frequency feature statistics and SHapley Additive exPlanations (SHAP) interpretability analysis, critical design features and their contribution mechanisms to multi-dimensional sustainability are elucidated. A widely distributed Pareto front is obtained, and statistical analysis reveals that recyclable materials (D13) and modular design (D1) constitute core features simultaneously enhancing sustainability across all three dimensions. Furthermore, reinforced frames (D5) and anti-fouling/antibacterial materials (D15) are found to contribute significantly to the safety and health aspect within the social dimension. The values for high-performance materials (D14) and multi-functionality (D6) exhibit considerable variability, necessitating trade-offs contingent upon specific design objectives. Representative optimal solutions from the Pareto front are selected as quantitative benchmarks, and the optimized feature parameters are hierarchically translated into material selection, structural schemes, and functional configurations, thereby completing the conceptual design of an automobile seat. The proposed BO–RF–NSGA-II hybrid framework is demonstrated to systematically address multi-objective conflicts inherent in sustainable design, thereby providing quantifiable and interpretable decision support for designers and effectively advancing the synergistic innovation of green, human-centric, and economically viable automobile seat development.</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

Data-driven sustainable design of automobile seats: integrating surrogate modeling and multi-objective optimization

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月29日

查看全文
  • Shiwen Huang,
  • Shibo Xu,
  • Xuhui Chen

摘要

To systematically enhance sustainability in automobile seat design and to holistically address the multi-dimensional requirements of the economic, environmental, and social dimensions under the Triple Bottom Line (TBL) framework, a data-driven design methodology integrating surrogate modeling with multi-objective evolutionary algorithms is proposed. First, user core needs are quantified and translated into 15 design features through SET (Social–Economic–Technical) factor analysis and the Fuzzy Analytic Hierarchy Process (FAHP). Subsequently, 10 key sustainability indicators are identified based on the TBL framework and categorized into economic, environmental, and social dimensions. Utilizing a dataset of 800 expert–student scoring records, global optimization of Random Forest (RF) hyperparameters is performed via Bayesian Optimization (BO), thereby constructing a high-accuracy surrogate model capable of predicting scores for any combination of design features across the identified sustainability indicators. The weighted composite scores for the three dimensions are adopted as optimization objectives, and the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed to explore Pareto-optimal solution sets for design feature configurations. Through high-frequency feature statistics and SHapley Additive exPlanations (SHAP) interpretability analysis, critical design features and their contribution mechanisms to multi-dimensional sustainability are elucidated. A widely distributed Pareto front is obtained, and statistical analysis reveals that recyclable materials (D13) and modular design (D1) constitute core features simultaneously enhancing sustainability across all three dimensions. Furthermore, reinforced frames (D5) and anti-fouling/antibacterial materials (D15) are found to contribute significantly to the safety and health aspect within the social dimension. The values for high-performance materials (D14) and multi-functionality (D6) exhibit considerable variability, necessitating trade-offs contingent upon specific design objectives. Representative optimal solutions from the Pareto front are selected as quantitative benchmarks, and the optimized feature parameters are hierarchically translated into material selection, structural schemes, and functional configurations, thereby completing the conceptual design of an automobile seat. The proposed BO–RF–NSGA-II hybrid framework is demonstrated to systematically address multi-objective conflicts inherent in sustainable design, thereby providing quantifiable and interpretable decision support for designers and effectively advancing the synergistic innovation of green, human-centric, and economically viable automobile seat development.