铁路道砟形态特征的统计分析与几何重构
作者:
作者单位:

1.同济大学 上海市轨道交通结构耐久与系统安全重点实验室,上海 201804;2.同济大学 道路与交通工程教育部重点实验室,上海 201804;3.上海公路桥梁(集团)有限公司,上海 200433

作者简介:

肖军华(1980—),男,教授,博士生导师,工学博士,主要研究方向为轨道交通土木结构。E-mail: jhxiao@tongji.edu.cn

通讯作者:

基金项目:

国家自然科学基金(51678447)


Statistical Analysis and Reconstruction of Morphological Characteristics of Railway Ballast
Author:
Affiliation:

1.Key Laboratory of Rail Infrastructure Durability and System Safety, Tongji University, Shanghai 201804, China;2.Key Laboratory of Road and Traffic Engineering of Ministry of Education, Tongji University, Shanghai 201804, China;3.Shanghai Road and Bridge (Group) Co. Ltd., Shanghai 200433, China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
    摘要:

    道砟颗粒几何形态特征对其力学特性影响显著。为了量化研究铁路道砟形态特征,以真实的铁路一级碎石道砟颗粒为例,采用3D激光扫描获取道砟颗粒点云数据,引入道砟整体形态特征指标(长轴、中轴、短轴、球度指数),提出道砟局部形态特征指标(曲率指数),统计并建立上述整体和局部形态特征指标的概率密度分布函数。在此基础上,基于本征正交分解(POD)及径向基(RBF)神经网络提出了基于颗粒形态指标概率密度分布的道砟样本的重生成算法,重构道砟颗粒样本库。采用上述重生成算法分别重构了600及4 000个颗粒道砟,结果表明:重构道砟样本形态特征指标的统计分布结果均与真实扫描试样结果接近,证明该方法能够快速实现基于颗粒形态指标概率密度分布的任意数量道砟样本的建立。

    Abstract:

    The geometrical shape characteristics of the particles have a significant influence on the mechanical properties of the ballast. To quantitatively study the morphological characteristics of railway ballast, this paper took the first-order ballast particles of the real scan as an example, used 3D laser scanning to obtain the point cloud data of the ballast particles, and introduced the description indices of overall morphological characteristics (long axis, middle axis, short axis, sphericity index), and proposed the local morphological characteristics index (curvature index) of the ballast granules. The probability density distributions of above-mentioned overall and local morphological characteristic indicators were established. On this basis, based on the intrinsic orthogonal decomposition (POD) and radial basis function (RBF) neural networks, a regenerative algorithm based on the probability density distribution of particle morphological indicators was proposed to reconstruct the ballast particle sample library. The above-mentioned re-generation algorithm reconstructed 600 and 4 000 particle turnouts, respectively. The results show that the statistical distribution of the morphological characteristics of the rebuilt ballast sample were close to those of the scanned samples, indicating that the method can quickly establish any number of turnout samples based on the probability density distribution of the particle morphological indicators.

    表 1 不同维数基向量重构的道砟颗粒形态及所表征的形态信息(以No.3-149#颗粒为例)Table 1
    图1 道砟3D激光扫描试验级配曲线Fig.1 Gradating curves of 3D laser scanning test for ballast
    图2 3D激光扫描Fig.2 3D laser scanning of ballast particles
    图3 道砟颗粒长轴、中轴、短轴计算示意图Fig.3 Schematic diagram of long axis, middle axis and short axis of ballast
    图4 道砟整体形态特征参数统计分布直方图及拟合概率密度曲线与函数Fig.4 Histogram, fitting probability density curves and function of the statistical of ballast morphological feature indices
    图5 离散点云局部曲率对道砟棱角的识别Fig.5 Identification of the corners and corners of the local curvature of discrete point clouds
    图6 道砟颗粒点云坐标标准化处理过程Fig.6 Normalization process of ballast particle point cloud coordinate
    图7 No.3-149#道砟颗粒前20%曲率对应的坐标点Fig.7 No.3-149# coordinate point corresponding to the first 20% curvature of the ballast particles
    图8 整体形态指标相近情况下颗粒CI与AI值对比Fig.8 Comparison of particle CI and AI values in the case of similar overall morphological indicators
    图9 曲率指数统计分布直方图及拟合概率密度曲线与函数Fig.9 Curvature index statistical distribution histogram and fitting probability density curve and function
    图10 各个方向上平均径向距离组成的颗粒Fig.10 Particles consisting of average radial distances in all directions
    图11 不同维数基向量所描述的颗粒形态特征的误差Fig.11 Errors of particle morphological features described by different dimensional basis vectors
    图12 不同级配原始颗粒与重构颗粒对比Fig.12 Comparison of original particles and reconstructed particles with different gradation
    图13 道砟样本重生成流程Fig.13 Progress of ballast sample regenerating
    图14 重生成不同颗粒数量的道砟样本特征参数分布对比Fig.14 Comparison of eigenvalue distributions of ballast samples regenerated with different particle numbers
    表 2 扫描样本与重生成样本形态特征指标分布的拟合概率密度函数中变量的平均值和标准差Table 2
    参考文献
    相似文献
    引证文献
引用本文

肖军华,郭佳奇,张德,薛立华.铁路道砟形态特征的统计分析与几何重构[J].同济大学学报(自然科学版),2020,48(12):1758~1769

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
历史
  • 收稿日期:2020-05-04
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2020-12-31