Scholarlink

<p>Traditional Chinese timber structures (TCTSs) exhibit significant seismic vulnerability. The installation of friction dampers at their mortise-tenon joints is an effective retrofit strategy, enhancing energy dissipation and reducing seismic responses. However, the optimization design of such dampers presents considerable challenges due to their multi-parameter nature combined with the pronounced sensitivity of TCTSs to stiffness variations. To address this issue, an optimization framework for the optimal design of friction dampers in TCTSs was proposed. The main steps include determining optimization parameters and ranges, establishing a sample database, training and optimizing predictive surrogate models (multi-layer perceptron, MLP), optimizing the multi-objective problem through Non-dominated Sorting Genetic Algorithm-II (NSGA-II), and identifying the optimal parameter combination using the entropy-weighted technique for order preference by similarity to ideal solution (TOPSIS) method. A TCTS retrofitted with replaceable displacement-amplification rotary friction dampers (RDARFDs) served as a case study, and the detailed Extended Discrete Element Method (EDEM) model was established. The results demonstrate the framework’s efficacy. The framework accurately predicts peak structural responses while effectively balancing displacement and acceleration control objectives. For the maximum inter-story drift ratio, the surrogate model achieves a coefficient of determination (R2) of 0.9924 and reduces the mean squared error (MSE) by 46.7% relative to the suboptimal model. Similarly, for peak floor acceleration, it yields an R<sup>2</sup> of 0.9568 with a corresponding MSE reduction of 15.8%. The optimal friction damper configuration was identified as [<i>µ</i> = 0.302, <i>ε</i> = 102 <i>µε</i>, <i>λ</i> = 0.566]. This design reduces the maximum inter-story drift ratio by 46.36% (from 1.45% to 0.78%) while increasing the peak floor acceleration by only 11.72% (from 1.45&#xa0;m/s<sup>2</sup> to 1.66&#xa0;m/s<sup>2</sup>), effectively controlling displacement without significantly amplifying acceleration. Furthermore, the proposed framework achieved a computational speedup of approximately 57.68 times, making the optimization process highly tractable.</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

Machine learning-aided optimization framework for friction dampers in traditional Chinese timber structures and case study evaluation

  • 期刊论文

  • 发表日期：

    2026年7月20日

查看全文
  • Leilei Liu,
  • Xicheng Zhang,
  • Yiwen Wu,
  • Jiajun Xu,
  • Yao Cui

摘要

Traditional Chinese timber structures (TCTSs) exhibit significant seismic vulnerability. The installation of friction dampers at their mortise-tenon joints is an effective retrofit strategy, enhancing energy dissipation and reducing seismic responses. However, the optimization design of such dampers presents considerable challenges due to their multi-parameter nature combined with the pronounced sensitivity of TCTSs to stiffness variations. To address this issue, an optimization framework for the optimal design of friction dampers in TCTSs was proposed. The main steps include determining optimization parameters and ranges, establishing a sample database, training and optimizing predictive surrogate models (multi-layer perceptron, MLP), optimizing the multi-objective problem through Non-dominated Sorting Genetic Algorithm-II (NSGA-II), and identifying the optimal parameter combination using the entropy-weighted technique for order preference by similarity to ideal solution (TOPSIS) method. A TCTS retrofitted with replaceable displacement-amplification rotary friction dampers (RDARFDs) served as a case study, and the detailed Extended Discrete Element Method (EDEM) model was established. The results demonstrate the framework’s efficacy. The framework accurately predicts peak structural responses while effectively balancing displacement and acceleration control objectives. For the maximum inter-story drift ratio, the surrogate model achieves a coefficient of determination (R2) of 0.9924 and reduces the mean squared error (MSE) by 46.7% relative to the suboptimal model. Similarly, for peak floor acceleration, it yields an R2 of 0.9568 with a corresponding MSE reduction of 15.8%. The optimal friction damper configuration was identified as [µ = 0.302, ε = 102 µε, λ = 0.566]. This design reduces the maximum inter-story drift ratio by 46.36% (from 1.45% to 0.78%) while increasing the peak floor acceleration by only 11.72% (from 1.45 m/s2 to 1.66 m/s2), effectively controlling displacement without significantly amplifying acceleration. Furthermore, the proposed framework achieved a computational speedup of approximately 57.68 times, making the optimization process highly tractable.