Scholarlink

<p>To address the issues of inaccurate modeling of nonlinear mechanical responses in magnetorheological seismic isolation bearings under complex seismic motions and inefficient analysis of actual bridge damping performance, this study proposes a hybrid analysis framework (BWMLA) that integrates the Bouc-Wen-Kelvin parallel physical model with long short-term memory (LSTM) network-based temporal prediction. The core innovation of this work lies in constructing a Bouc-Wen-Kelvin parallel physical characterization model. By employing a phased collaborative optimization approach using genetic algorithms and Bayesian optimization for both physical parameters and LSTM network hyperparameters, it achieves a deep integration of physical mechanisms and data-driven insights. Field validation on a suspension bridge with a main span of 385&#xa0;m demonstrated that the proposed framework achieved a root mean square error (RMSE) of 9.2 kN and a frequency-domain energy storage modulus error (Δk) of 5.1% for hysteretic force modeling, significantly outperforming the reference model. In the actual bridge testing, the RMSE between the predicted and measured values for structural deformation displacement and acceleration were 19.7&#xa0;mm and 0.30&#xa0;m/s<sup>2</sup>, respectively, with the coefficient of determination (R<sup>2</sup>) for the response spectrum reaching 0.91. This model provides a highly reliable tool for evaluating the performance of real-bridge vibration isolation systems and implementing adaptive control. Its high precision and efficiency enable direct application in guiding the optimized design of real-bridge isolation systems and the development of real-time seismic monitoring systems.</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

Magneto-rheological isolation bearings vibration damping performance in an actual bridge based on the Bouc-Wen-Kelvin parallel model and second-order optimized LSTM: a case study of a 385m suspension bridge

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月14日

查看全文
  • Haijun Cui

摘要

To address the issues of inaccurate modeling of nonlinear mechanical responses in magnetorheological seismic isolation bearings under complex seismic motions and inefficient analysis of actual bridge damping performance, this study proposes a hybrid analysis framework (BWMLA) that integrates the Bouc-Wen-Kelvin parallel physical model with long short-term memory (LSTM) network-based temporal prediction. The core innovation of this work lies in constructing a Bouc-Wen-Kelvin parallel physical characterization model. By employing a phased collaborative optimization approach using genetic algorithms and Bayesian optimization for both physical parameters and LSTM network hyperparameters, it achieves a deep integration of physical mechanisms and data-driven insights. Field validation on a suspension bridge with a main span of 385 m demonstrated that the proposed framework achieved a root mean square error (RMSE) of 9.2 kN and a frequency-domain energy storage modulus error (Δk) of 5.1% for hysteretic force modeling, significantly outperforming the reference model. In the actual bridge testing, the RMSE between the predicted and measured values for structural deformation displacement and acceleration were 19.7 mm and 0.30 m/s2, respectively, with the coefficient of determination (R2) for the response spectrum reaching 0.91. This model provides a highly reliable tool for evaluating the performance of real-bridge vibration isolation systems and implementing adaptive control. Its high precision and efficiency enable direct application in guiding the optimized design of real-bridge isolation systems and the development of real-time seismic monitoring systems.