摘要
针对重载铁路钢轨表面伤损识别问题,提出一种基于钢轨廓形数据的钢轨表面伤损辨识方法.首先,提出基于统计特征与多尺度排列熵相结合的混合特征提取方法,解决单一特征提取方法所提取的伤损特征表征不够全面的问题;其次,提出基于ReliefF和改进蚁群算法的二阶特征选择方法,实现钢轨伤损有效特征的快速选择,得到最佳特征集合;最后,采用基于径向基核函数(RBF)的支持向量机(SVM)实现钢轨表面伤损辨识.依托株洲中车时代电气股份有限公司采集的7种常见工况下共计3500组廓形数据开展实验,结果表明:相比于单一特征提取方法及单一特征选择方法,所提方法能够进一步提升重载铁路钢轨表面伤损识别准确率,可达99.43%.
To address the problem of identifying surface damage on heavy haul railway rails,a rail surface damage identification method based on rail profile data was proposed.Firstly,the hybrid feature extraction method combining statistical features and multi-scale permutation entropy was proposed to solve the problem that the single feature extraction method is not comprehensive in the representation of damaged features.Secondly,the two-stage feature selection method based on ReliefF and improved ant colony algorithm was proposed to realize the quick selection of effective features of rail damage and obtain the best feature set.Finally,the support vector machine(SVM)based on the radial basis kernel function(RBF)was used to identify the rail surface damage.Experiments were conducted based on 3500 sets of profile data collected from CRRC Zhuzhou Institute Co.Ltd..under 7 common working conditions.Results shows that compared with single feature extraction methods and single feature selection methods,the proposed method can further improve the surface damage identification accuracy of heavy haul railway rails,reaching 99.43%.
作者
孙永奎
孙思琦
曹源
宿帅
SUN Yongkui;SUN Siqi;CAO Yuan;SU Shuai(School of Automation and Intelligence,Beijing Jiaotong University,Beijing 100044,China;National Engineering Research Center of Rail Transportation Operation and Control System,Beijing Jiaotong University,Beijing 100044,China)
出处
《华中科技大学学报(自然科学版)》
北大核心
2025年第8期90-97,共8页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家重点研发计划项目资助项目(2021YFF0501102)
国家自然科学基金资助项目(52202392,U2368202,U1934219)
中央高校基本科研业务费专项资金资助项目(2024JBMC027).
关键词
钢轨表面伤损辨识
廓形数据
二阶特征选择
改进蚁群算法
支持向量机
rail surface damage identification
profile data
two-stage feature selection
improved ant colony algorithm
support vector machine
