基于主元分析新统计量的多元统计过程监控(英文)

Multivariate statistical process monitoring based on new statistics of principal component analysis

针对传统的主元分析(PCA)的T^2和平方预测误差(SPE)检验所提供的信息并不一致的缺陷,提出了一种改进的PCA方法。该方法采用主元相关变量残差(PVR)和一般变量残差(CVR)统计量代替SPE统计量用于过程监测。将此改进的PCA方法应用到双效蒸发过程的仿真监测,与传统的PCA方法相比,新PCA方法能够有效地识别正常工况改变与过T^2程故障引起的图变化,避免了SPE统计量的保守性,能够提供更详细的过程变化信息,提高了对过程变化的分析与诊断能力。

The information provided by T^2 and squared prediction error（SPE） test of principal component analysis（PCA） is not corresponding. An improved PCA was proposed which uses principal-component-related variable residual statistic and common variable residual statistic to replace SPE statistic. Then a simulated double-effect evaporator was monitored by using the proposed method and comparisons with the conventional PCA are made. The simulation result shows that the root cause that violates the T^2 test but still satisfies SPE test can be unambiguously identified and the improved PCA can avoid the conservation of SPE statistical test and provide more explicit information about the process conditions. So the improved PCA has an enhanced fault diagnosing performance.