基于时间序列ARMA模型的振动故障预测  

Vibration Faults Prediction Based on Time Series Autoregressive Moving Average(ARMA) Model

作　　者：刘颖[1] 严军[1] 

机构地区：[1]西北师范大学化学化工学院,兰州730070

出　　处：《化工自动化及仪表》2011年第7期841-843,共3页Control and Instruments in Chemical Industry

摘　　要：运用自回归滑动平均(ARMA)模型和聚类分析方法确定参考样本和故障样本的特征向量,通过特征向量的距离识别故障类型。根据汽轮机典型故障构造模拟信号,建立其ARMA预测模型,通过聚类分析得出标准信号及待测信号的特征向量。经验证,基于ARMA预测模型和聚类分析的方法能够正确识别故障类型。By making use of autoregressive moving average（ARMA） model and clustering method,the characteristic vectors of reference samples and fault samples were determined to identify the fault type according to the feature vector distance;and basing on the analog signals from the typical turbine faults,their ARMA prediction models were built to work out the characteristic vector of reference signals and test signals.The testing result proves the success of this method.

关 键 词：汽轮机 故障 ARMA模型 时间序列 聚类分析 

分 类 号：TP306.3[自动化与计算机技术—计算机系统结构]