基于相对主元分析的故障检测与诊断方法

Relative PCA-based Fault Detection and Diagnosis Method

针对传统主元分析(PCA)在实际监控过程中存在的问题,文中提出相对主元(RPC)的概念,并建立了一种相对主元分析方法(RPCA)。RPCA在有效处理丢失数据及异常点问题的同时,能够根据各分量在系统中的不同重要性,赋以相应的权值,从而达到建立相对精确的RPC模型进行故障检测和诊断的目的。与传统方法相比,RPCA可以克服其在实际故障检测时能力的不足。给出的计算机仿真实验证明了该方法的有效性和实用性。

Aiming at the problems happened in the practical monitoring processing for traditional principle component analysis （PCA）, the concept of relative principle component （RPC） is put forward as well as method of relative principle component analysis （RPCA）, which can deal with the phenomenon of missing data and outlier effectively, endow with corresponding weights according to different importance of each variable in system and thereby built the relative accurate model to detect and diagnose fault. Comparing with the traditional method, RPCA can overcome some disadvantages of which for practical fault detection. A simulation has been given to demonstrate the effectiveness and practicability of the algorithm proposed.