中文摘要：

由于基于主元分析（Principal Component Analysis，PCA）的统计监控方法没有利用过程机理模型（First Principle Model）信息，因此在一定程度上限制了其故障诊断能力的发展。本文基于PCA的框架，采用故障子空间对故障进行描述，在PCA监测模型的基础之上，分析了主元空间和残差空间的故障可检测性问题，获得了故障可检测性的必要充分理论条件。通过对双效蒸发过程的仿真监测，证实了所获理论结果的有效性，表明了通过计算临界故障幅值就可事先对故障集内各故障的检测结果作定量的分析，从而事先了解各故障在PCA下的检测结果。

英文摘要：

The most significant advantage of principal component analysis （PCA） is that no precise process model is needed. Nevertheless, because the first principle model information is not utilized by statistical monitoring approach based on PCA, the development of capacity for fault diagnosis is restricted to some extent. Based on the frame of PCA, using the fault subspace to depict fault, the problem about fault detectability in the component space and the residual space is analyzed based on PCA monitoring model. The necessary and sufficient theoretical conditions of fault detectability are obtained. Through the simulation monitoring of double-effect evaporator process, these results show that the acquired theoretical results are effective and in order to advance understanding of detection results in the PCA under various fault, prior to detection of the fault in the set of results for quantitative analysis can be done through caleulating the critical fault magnitude.