基于K-均值聚类分析的城市道路汽车行驶工况构建方法研究

摘要

以实时采集的乘用车行驶数据为数据源,进行了城市道路汽车行驶工况构建方法的研究.分别运用运动学片段分析法、主成分分析法和K均值聚类分析法对实测数据进行降维和分类,提出以Silhouette函数实现对聚类结果的筛选,以减少人为选择的误差,并根据聚类中心的大小筛选所需运动学片段构建候选工况.在目标代表工况的遴选方面,提出了综合6个特征参数和最大SAFD差异值的评价标准.最后通过试验验证了该行驶工况构建方法的有效性和精确性.%The construction method of city road driving cycle was studied with the real-time collected passenger vehicle driving data as data source. The kinematics fragment analysis, principal component analysis and K-means clustering analysis were applied separately for dimensionality reduction of the measured parameters and classification, then Silhouette Function was proposed to screen the clustering results, to reduce artificial selection error, and the required kinematic fragment candidate cycle was selected according to size of the clustering center. In term of choice of typical driving cycle, an evaluation criterion that integrated six characteristic parameters and the maximum SAFD difference value was proposed. Finally test proved validity and accuracy of the construction method of the driving cycle.