密度偏差抽样算法的设备样本点选择

打开文本图片集

中图分类号：TP391.4 文献标志码：B 文章编号：1671-5276（2025）05-0057-05

Equipment Sample Point Selection for Density Biasd Sampling Algorithm

CHEN Shuang，WU Tiejun

CollegeofMechanicalandElectricalEnginering，NanjingUniversityofAeronauticsandAstronautics，Nanjing2oo6，China）

Abstract：A devicetypical samplepointselectionmethod basedonvariable grid and FCMdensity bias sampling is proposed to addressthe problems of enormous consumption of time and space resources in data一driven device fault diagnosis bydirect calculation.The dataset is divided into several gridsof different sizesbyavariable gridpartitioning method，the expected number of samples in each grid are calculated based on the sampling probabilityof data points within each grid unit，andFCMclustering is performed insidethe grid toobtain the numberof target sample points.The experimentalresults show thattheproposedalgorithmnotonly ensures the integrityoftheoriginal dataset，but also maintains relatively accurate clustering efect，which lays foundation for the subsequent recognitionof equipment operating conditions.

Keywords：density biased sampling；fuzzy C- means;variable grid;typical sample points of equipment

0 引言

面对海量和高维的设备运行数据，如果直接聚类分析来获取其运行状态，会使得运行的效率与聚类的质量均无法得到较好的保证，且会有运行时间过长，甚至内存溢出导致无法计算的问题[1]。（剩余5828字）