基于融合评价指标的k-means聚类算法的地铁车轮踏面磨耗分析

Metro Wheel Tread Wear Analysis Based on k-means Clustering Algorithm Based on Fusion Evaluation Index

采用聚类思想对某地铁线路大量车轮踏面磨耗数据进行特征提取,对其磨耗特征进一步分析,针对聚类参数转换方法对不同聚类效果评价指标造成的影响开展研究,提出基于融合评价指标的k-means均值聚类方法,解决利用聚类模型确定聚类数时,主观因素对聚类效果的干扰。结果表明:以轮缘厚度、轮缘高度及轮缘综合值作为聚类特征,以融合评价指标作为最佳聚类数的选择依据,采用相应聚类特征参数的方差对其加权方法进行聚类特征变换,能得到较好的聚类效果;将地铁车轮踏面聚成5类,采用均值的方法划分出5类典型磨耗廓形,并基于不同时间节点的车轮外形数据进行计算分析,进一步验证了该聚类方法的有效性,为地铁车轮踏面经济镟修策略提供了参考。

The clustering idea is used to extract the features of a large number of wheel tread wear data of a subway line,the wear characteristics are further analyzed,the impact of the clustering parameter conversion method on different clustering effect evaluation indicators are studied,and a k-means mean clustering method based on fusion evaluation index is finally proposed,which solves the interference of subjective factors on the clustering effect of the clustering model when determining the number of clusters.The results show that the cluster feature transformation of the weighting method is carried out by using the variance of the corresponding clustering feature parameters;the subway wheel tread is clustered into five categories,the average method is used to divide the five typical wear profiles,and the effectiveness of the clustering method is further verified by calculation and analysis based on the wheel shape data at different time nodes.It provides reference for the economic repair strategy of metro wheel tread.