针对转辙机退化阶段难以划分的问题,提出一种基于多维特征融合的道岔转辙机退化状态识别方法.首先,提取了 S700K转辙机退化功率数据的时域、频域、时频域多域特征;其次,通过核主成分分析(Kernel Principal Components Analysis,KPCA)进行特征融合,获得表征道岔转辙机运行状态的特征向量,构建转辙机退化性能指标;再次,采用K-medoids聚类算法对道岔转辙机性能退化状态进行阶段划分,识别不同的退化状态;最后,选用轮廓系数、分类系数、平均模糊熵对聚类效果进行综合评价,并与模糊C均值聚类(Fuzzy C-Means Clustering,FCM)和古斯塔夫森-凯塞尔(Gustafson Kessel,GK)聚类算法进行比较.研究结果表明,融合特征聚类后的综合评价指标高于单一特征,更能够体现道岔转辙机退化过程中的细节,K-medoids聚类效果明显,模型的准确率达到96.3％,能够对道岔转辙机性能退化阶段进行准确的划分,为铁路现场道岔智能运维提供理论支持.
Identifying turnout switch machine degradation states via multi-feature fusion and cluster analysis
Addressing the challenge of accurately categorizing rutting machine degradation stages,this paper proposes a method for identifying the degradation state of turnout rutting machines through multidimensional feature fusion.Initially,time-domain,frequency-domain,and time-frequency-domain features are extracted from the degraded power data of S700K rutters.To address the modal aliasing issue encountered during feature extraction in the time-frequency domain using Empirical Mode Decomposition(EMD),we introduce Variational Mode Decomposition(VMD)for signal decomposition.Following decomposition,the power signal of the rutters undergoes further analysis.We utilize VMD to decompose the power signal of rutters,resulting in distinct Intrinsic Mode Functions(IMF)components.We then calculate the fuzzy entropy of each IMF as a time-frequency domain feature.Next,employing Kernel Principal Components Analysis(KPCA),we fuse these multi-domain features to generate feature vectors that represent the operational state of the turnout rutting machine.Subsequently,we construct degradation performance indexes for the rutting machine.Next,we apply the K-medoids clustering algorithm to classify the stages of rutting machine performance degradation.This process involves determining the clustering center for each degradation state and establishing a rutting machine degradation state identification model based on the affiliation between data samples and clustering centers to assign them to their respective stages.Finally,we conduct a comprehensive evaluation of the clustering effectiveness by selecting contour coefficients,classification coefficients,and average fuzzy entropy.These metrics are compared with those obtained from the Fuzzy C-means Clustering(FCM)and the Gustafson-Kessel(GK)clustering algorithms.The findings reveal that the integrated evaluation index following feature fusion clustering surpasses that of using single features alone,offering a more detailed depiction of the turnout rutting machine degradation process.In comparison to the FCM and GK clustering algorithms,the K-medoids algorithm demonstrates a noticeable improvement in clustering effectiveness,achieving a model accuracy of 96.3％.This enables precise delineation of performance degradation stages in turnout machines,thus furnishing theoretical underpinnings for intelligent on-site operation and maintenance of railway turnouts.
万方数据
维普
