摘要： 高速运行的离心机设备,其振动状态检测数据通常呈现出明显的非线性、正态分布和大样本的特征,数据波动的随机性使得其趋势特征难以捕捉。为此,提出一种新的时间序列模式分类方法。采集离心机设备运行状态的振动信号时间序列进行分析,根据对称原理提取序列数据的分位数,构建特征向量,采用欧氏距离函数构建相似性度量,建立模式分类的判定依据,使用k-means 分类算法实现状态模式的自动分类。仿真结果表明,该方法能有效区分离心机设备运行中空载和负载的模式状态,且比传统的小波分析模式分类方法更加准确。
关键词:
分位数,
时间序列,
模式分类,
离心机,
振动信号,
小波
Abstract: The vibration state detection data from the centrifuge equipment in high-speed operation usually presents
obvious nonlinearity,normal distribution and the characteristics of large sample,and random fluctuations in the data make it difficult to capture the trend characteristics. In this paper,time sequence theory is used to analyze the vibration signal data gathered from the running centrifuge equipment. It uses the quantile of the sequence data to build the feature vector according to the symmetrical principle,and introduces the Euclidean distance function to construct similarity measure,and then sets up the decision basis for pattern classification, realizes the pattern classification employing the k-means classification algorithm. Simulation results show that this method can effectively distinguish the partial load state and noload state of the centrifuge equipment,which is more accurate than that of wavelet analysis method.
Key words:
quantile,
time sequence,
pattern classification,
centrifugal machine,
vibration signal,
wavelet
中图分类号: