摘要
针对目前地铁构架应力谱编制中应力分级划分不合理导致评估结果与实际疲劳损伤存在较大差异的问题,研究基于模糊聚类的高表征度地铁构架应力谱编制方法。首先,采用模糊聚类分析方法,对不同应力循环造成的损伤进行聚类;然后,根据不同特征应力造成的损伤隶属度进行级数确定,对由于低中应力区的应力大小和次数变化较大,使得相邻应力级平顺性较差的问题,采用神经网络曲线拟合进行平滑处理,而对由于高应力区存在次数为零的应力级,且整体次数较少,使得数据本身的数理统计规律性较弱的问题,则采用二维核密度估计进行非参数统计;最后,基于损伤一致性原则,对分类结果中各级等效应力进行优化,完成高表征度应力谱的编制。以某地铁转向架构架制动吊座与横梁连接处测试数据为例,与传统雨流计数编谱方法对比,验证本文方法。结果表明:采用模糊聚类和应力分区统计的编谱方法,不仅可以准确反映应力的分布规律,真实表征应力作用所产生的疲劳特性,而且外推结果可有效克服有限数据样本采集不足引起的误差,有效模拟了结构实际受载状况,提高了所编制应力谱在构架疲劳分析方面的准确性,并可为其他应力谱编制提供参考。
To address the problem of large differences between the evaluation results and the practical fatigue damage due to unreasonable stress classification and gradation in the compilation of the stress spectrum for subway frames, a compilation method for the stress spectrum of high-characteristic subway frames based on fuzzy clustering is studied. Firstly, the fuzzy clustering analysis method is applied to cluster the damage caused by different stress cycles. Then, the level is determined according to the degree of damage membership caused by stresses with different characteristics. The changes of stress magnitudes and numbers in the low and the medium stress areas appear to be great, resulting in poor regularity among adjacent stress levels. To address this problem, this research adopts neural network curve fitting for regularization, while for the existence of stress level with constant stress and the fewer overall change numbers in high stress areas, and the data with weak properties regarding the principle of mathematical statistics, a two-dimensional kernel density estimation(KDE) is used for non-parametric statistics. Finally, based on the principle of damage consistency, the equivalent stresses at all levels in the classification results are optimized to complete the compilation of stress spectra with high degrees of characterization. Taking the test data of the joint of the brake hanger and the beam of a subway bogie frame as an example, the proposed method in this paper is verified by comparing it with the traditional rainflow counting spectrum compilation method. The results show that, the spectrum compilation method using fuzzy clustering and stress partitioning statistics can not only reflect the distribution law of stress and truly characterize the fatigue properties produced by the stress action, but also that the extrapolation results can overcome the errors caused by the insufficient collection of limited data samples. Thus the proposed method stimulates the load condition of the structure with greater
作者
薛海
胡李军
李强
XUE Hai;HU Lijun;LI Qiang(School of Mechatronic Engineering,Lanzhou Jiaotong University,Lanzhou Gansu 730070,China;Engineering Research Center of Structure Reliability and Operation Measurement Technology of Rail Guided Vehicles of Ministry of Education,Beijing Jiaotong University,Beijing 100044,China)
出处
《中国铁道科学》
EI
CAS
CSCD
北大核心
2022年第5期102-110,共9页
China Railway Science
基金
甘肃省科技计划项目(20JR5RA412)
北京交通大学轨道车辆结构可靠性与运用检测技术教育部工程研究中心开放课题(219008)
兰州交通大学-天津大学联合创新基金资助项目(2020058)
兰州交通大学天佑创新团队项目(TY202006)。
关键词
应力谱
疲劳损伤
模糊聚类
神经网络
二维核密度估计
雨流计数
Stress spectrum
Fatigue damage
Fuzzy clustering
Neural network
Two-dimensional kernel density estimation
Rainflow counting
