• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2026, Vol. 62 ›› Issue (15): 183-194.doi: 10.3901/JME.260826

• 特邀专栏:大模型时代的智能设计与制造 • 上一篇    

扫码分享

基于PSO-BP和NSGA-Ⅱ的地铁嵌入式轨道结构优化研究

张洪1,2, 杜显德2, 刘林芽2,3, 张鹏飞1,2, 成功2, 王怀东4, 王玮5   

  1. 1. 华东交通大学综合立体交通信息感知与融合江西省重点实验室 南昌 330013;
    2. 华东交通大学交通运输工程学院 南昌 330013;
    3. 萍乡学院 萍乡 337055;
    4. 中铁九局集团有限公司 沈阳 110051;
    5. 极端环境岩土和隧道工程智能建养全国重点实验室(铁一院) 西安 710043
  • 收稿日期:2025-08-11 修回日期:2026-01-22 发布日期:2026-10-13
  • 作者简介:张洪,男,1989年出生,博士,副教授,硕士研究生导师。主要研究方向为线路智能优化;铁路减振降噪。E-mail:hongzhang@ecjtu.edu.cn;杜显德,男,1999年出生,硕士研究生。主要研究方向为铁路减振降噪。E-mail:13155899161@163.com;刘林芽,男,1973年出生,博士,教授,博士、硕士研究生导师。主要研究方向为轨道交通振动与噪声;交通基础设施智能建造;轨道交通基础设施安全与维护。E-mail:lly1949@163.com;张鹏飞,男,1975年出生,博士,教授,博士、硕士研究生导师。主要研究方向为轨道交通环境振动预测、评价及控制;轨道结构及轨道动力学;桥上无砟轨道无缝线路设计理论及关键技术。E-mail:zhangpf4236@163.com;成功,男,1990年出生,博士,讲师,硕士研究生导师。主要研究方向为轨道交通振动与噪声。E-mail:cxglam@163.com;王怀东,男,1977年出生,学士,高级工程师。主要研究方向为铁路轨道智能建造及检测。E-mail:718202272qq.com;王玮(通信作者),男,1983年出生,硕士研究生,高级工程师。主要研究方向为智能勘测;轨道交通智能化。E-mail:wangwei2121@qq.com
  • 基金资助:
    国家重点研发计划(2022YFB2602200)、国家自然科学基金(52178425, 52272348, 52308451)、中国博士后科学基金面上(2022M711137)、轨道交通工程信息化国家重点实验室开放课题(SKLK22-10)和江西省研究生创新专项基金(YC2024-S455)资助项目。

Optimization of Metro Embedded Track Structure Based on PSO-BP and NSGA-II

ZHANG Hong1,2, DU Xiande2, LIU Linya2,3, ZHANG Pengfei1,2, CHENG Gong2, WANG Huaidong4, WANG Wei5   

  1. 1. Jiangxi Provincial Key Laboratory of Comprehensive Stereoscopic Traffic Information Perception and Fusion, East China Jiaotong University, Nanchang 330013;
    2. School of Transportation Engineering, East China Jiaotong University, Nanchang 330013;
    3. Pingxiang University, Pingxiang 337055;
    4. China Railway No. 9 Engineering Group Co. Ltd., Shenyang 110051;
    5. State Key Laboratory of Intelligent Geotechnics and Tunnelling (FSDI), Xi'an 710043
  • Received:2025-08-11 Revised:2026-01-22 Published:2026-10-13

摘要: 嵌入式轨道相比扣件式轨道具有更好的轨道动态特性和车辆轨道相互作用性能,这使得嵌入式轨道拥有良好的减振降噪效果。然而,其复杂的结构对设计提出了更高的要求,如轨道结构参数取值不合理,将显著影响其动力学性能和减振降噪效果。为优化地铁嵌入式轨道结构参数,提出了一种基于PSO-BP神经网络和NSGA-Ⅱ的地铁嵌入式轨道结构优化方法,以轨道板长度、宽度、填充材料刚度、阻尼等9项参数为决策变量,钢轨垂向位移、钢轨垂向加速度、轨道板垂向位移等3项动力响应指标为目标函数,建立了嵌入式轨道多目标优化模型,并提出基于PSO-BP神经网络的嵌入式轨道动力响应预测方法,获取不同决策变量下的目标函数值。通过采用NSGA-Ⅱ对模型进行求解,实现嵌入式轨道结构优化,并引入Sobol方法分析不同结构参数对嵌入式轨道动力学性能的影响。结果表明:① PSO-BP神经网络可高效、准确预测嵌入式轨道动力响应,平均准确率可达97.15%;② 优化后嵌入式轨道的钢轨垂向位移、钢轨垂向加速度、轨道板垂向位移相比优化前分别减少了20%、18.7%、15.6%;③ 通过Sobol敏感性分析得出,填充材料垂向刚度、垂向阻尼、轨道板长度分别是影响钢轨垂向位移、钢轨垂向加速度和轨道板垂向位移的关键参数。本研究成果可为地铁嵌入式轨道结构设计提供参考。

关键词: 嵌入式轨道, PSO-BP神经网络, 多目标优化, NSGA-Ⅱ, Sobol敏感性分析

Abstract: Embedded tracks demonstrate superior dynamic characteristics and enhanced vehicle-track interaction performance compared to fastener-based tracks, providing them with excellent vibration reduction and noise mitigation effects. However, their complex structure imposes higher demands on design. Improper selection of structural parameters can significantly compromise the track’s dynamic performance and vibration reduction effects. To optimize the metro embedded track structure parameters, this study proposes an optimization method for metro embedded track structures based on the PSO-BP neural network and NSGA-II. A multi-objective optimization model is developed using nine structural parameters, including track slab length, track slab width, filler stiffness and damp, etc., as decision variables, and three dynamic performance indicators—rail vertical displacement, rail vertical acceleration, and track slab vertical displacement—as objective functions. A PSO-BP neural network-based prediction method is introduced to predict dynamic responses under different parameter settings, and NSGA-II is employed to optimize the model. Furthermore, the Sobol method is utilized to analyze the influence of various parameters on the dynamic performance of the embedded track. Results indicate that: (1) the PSO-BP neural network can accurately and efficiently predict the dynamic responses of embedded tracks, achieving an average accuracy of 97.15%; (2) after optimization, rail vertical displacement, rail vertical acceleration, and track slab vertical displacement are reduced by 20%, 18.7%, and 15.6%, respectively; (3) sensitivity analysis reveals that filler vertical stiffness, vertical damping, and track slab length are the key parameters influencing rail vertical displacement, rail vertical acceleration, and track slab vertical displacement, respectively. This research provides valuable references for the structural design of metro embedded tracks.

Key words: embedded track, PSO-BP neural network, multi objective optimization, NSGA-Ⅱ, Sobol sensitivity analysis

中图分类号: