| 基于特征熵权与样本加权的大跨桥梁多车道交通荷载聚类模型 |
| |
| 引用本文： | 郭雪莲,黄平明,韩万水,刘晓东,袁阳光,赵越.基于特征熵权与样本加权的大跨桥梁多车道交通荷载聚类模型[J].中国公路学报,2022,35(10):183-193. |
| |
| 作者姓名： | 郭雪莲 黄平明 韩万水 刘晓东 袁阳光 赵越 |
| |
| 作者单位： | 1. 长安大学 公路学院, 陕西 西安 710064;2. 西安建筑科技大学 土木工程学院, 陕西 西安 710055;3. 西安建筑科技大学 工程结构安全与耐久重点实验室, 陕西 西安 710055;4. 西安理工大学 土木建筑工程学院, 陕西 西安 710048 |
| |
| 基金项目： | 国家重点研发计划项目(2019YFB1600702)；国家自然科学基金项目(51878058,52008027)； 陕西省自然科学基础研究计划一般项目-青年项目(2020JQ-665,2019JZ-02) |
| |
| 摘 要： | 为满足交通流荷载作用下大跨桥梁结构评估的需要，研究了基于荷载参数特征的交通流状态划分方法。首先，基于实测交通流数据，按照车道属性统计分析得到交通流的单位小时特征参数样本，选择单位小时内车型比例、车头间距及交通流速度作为交通流状态划分的参考特征；其次，改进经典k-means聚类算法以增强其对高维、复杂交通流荷载分类的鲁棒性，即通过引入特征熵值来表征各特征参数对聚类效果的重要性，同时计算样本点与周围样本点的接近程度来赋予样本点权值，以削弱样本离散性对聚类质量的不利影响；最后，通过聚类算法得到11种具有不同参数特征的交通流荷载，分析了其作用下某大跨斜拉桥拉索应力响应及造成的疲劳损伤。结果表明：改进算法的聚类质量指标比原始k-means算法提高了40%以上，对交通流状态划分具有良好的适用性；通过算法得到的不同类别的交通流荷载的特征参数差异性明显，其占有率也大不相同，同一类别的交通流荷载各样本特征参数聚拢效果良好；同车道内不同类别的交通流荷载的拉索等效应力差别较大，其变异系数均在0.2以上，尤其在考虑了不同交通流荷载模型的占有率后，这种差异性进一步增大。上述结果表明该交通流荷载聚类与模拟方法是有效、准确的，对相关大跨桥梁结构安全及耐久性评估有一定的参考价值。
|
| 关 键 词： | 桥梁工程 大跨桥梁 结构评估 交通流荷载 交通流状态划分 聚类算法 |
| 收稿时间： | 2021-05-05 |
Multi-lane Traffic Load Clustering Model for Long-span Bridges Based on Weight of Feature Entropy and Sample |
| |
| GUO Xue-lian,HUANG Ping-ming,HAN Wan-shui,LIU Xiao-dong,YUAN Yang-guang,ZHAO Yue.Multi-lane Traffic Load Clustering Model for Long-span Bridges Based on Weight of Feature Entropy and Sample[J].China Journal of Highway and Transport,2022,35(10):183-193. |
| |
| Authors: | GUO Xue-lian HUANG Ping-ming HAN Wan-shui LIU Xiao-dong YUAN Yang-guang ZHAO Yue |
| |
| Affiliation: | 1. School of Highway, Chang'an University, Xi'an 710064, Shaanxi, China;2. College of Civil Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, Shaanxi, China;3. Key Lab of Engineering Structural Safety and Durability, Xi'an University of Architecture and Technology, Xi'an 710055, Shaanxi, China;4. School of Civil Engineering and Architecture, Xi'an University of Technology, Xi'an 710048, Shaanxi, China |
| |
| Abstract: | To meet the needs of structural evaluation of long-span bridge under traffic flow load, the method of the state classification of traffic flow load based on its feature of parameter was studied. Firstly, based on the measured traffic flow data, the samples of traffic flow parameters per hour were obtained according to the statistical analysis for different lanes. The vehicle type proportion, headway and traffic flow speed per hour were selected as the parameters for the classification of traffic flow load. Secondly, by improving the classical k-means clustering algorithm, its robustness to high-dimensional and complex traffic flow load classification was enhanced. The feature entropy was introduced to characterize the importance of each classification feature to the clustering effect. The proximity of the sample point to the surrounding sample points was calculated to assign weight to the sample point. Thereby, the unfavorable impact of the clustering quality resulted from the sample dispersion was weaken. In the last, 11 traffic flow loads with different parameter characteristics were obtained by the algorithm. The cable stress response and fatigue damage of a long-span cable-stayed bridge under these traffic loads were analyzed. The results show that the quality index of the improved algorithm clustering is more than 40% higher than that of the original k-means algorithm. At the same time, the improved algorithm has good applicability to the classification of traffic flow load. The characteristic parameters of different types of traffic flow load have obvious differences. The occupancy rate is also very different. In addition, the characteristic parameters of the samples of the same kind of traffic flow load have a good aggregation effect. The equivalent stress of the cables of different kinds of traffic flow loads in the same lane is quite different, and the coefficient of variation is above 0.2. Note that when considering the occupancy rates of different traffic flow models, this difference has further increased. The above results show that the clustering and simulation method of traffic flow load is effective and accurate, and can be the reference for the safety and durability evaluation of long-span bridge structures. |
| |
| Keywords: | bridge engineering long-span bridge structural evaluation traffic flow load state classification of traffic flow load clustering algorithm |
|
| 点击此处可从《中国公路学报》浏览原始摘要信息 |
|
点击此处可从《中国公路学报》下载全文 |
|
