基于LPP的工业过程故障检测

针对传统的降维算法在降维过程中存在着丢失数据的局部邻域信息的问题,一种基于局部保持投影(LPP)用于工业工程数据检测的方法被应用。LPP算法的思想是通过构造数据样本点之间的亲疏关系,并且在投影降维的同时保留数据样本点的这种局部邻域结构,从而保留数据的局部信息。论文将LPP算法与传统的降维算法-主元分析法(P CA)在田纳西-伊斯曼过程(T EP)仿真系统上进行检测对比,结果表明,LPP算法具有更加优越的检测性能。

Industrial Process Fault Detection Based on LPP

Aiming at the problem of losing local neighborhood information of data in traditional dimension reduction algo rithms,a manifold algorithm Local Preserving Projection(LPP)is adopted to detect industrial Engineering data.The idea of LPP al gorithm is to preserve the local neighborhood structure of data sample points while projecting dimensionality reduction,so as to pre serve the local information of data.This paper compares LPP algorithm with traditional dimension reduction algorithm-PCA in Ten nessee-Eastman Process(TEP)simulation system.The results show that LPP algorithm has better detection performance.