中文摘要：

摘要：基于主元分析（Principal Component Analysis，PCA）统计过程监控方法，由于其不需要数学模型，因此目前在过程监控领域获得了广泛应用，但这也限制了其在故障诊断方面的能力。针对此问题，本文从故障子空间与PCA监控模型的角度，利用故障重构技术，对基于PCA的严统计量进行重构，获得了主元子空间中严统计量的故障可重构性理论条件，提出了具体的故障识别指标和诊断算法，解决了基于主元子空间故障重构技术的故障诊断问题，弥补了Dunia等人的方法只在残差子空间中讨论故障重构与识别问题。通过对双效蒸发过程的仿真监控，表明了所获得的理论条件、故障识别指标和诊断算法能对传感器故障和过程故障进行有效地识别，证实了所获理论、识别指标和诊断算法的有效性。

英文摘要：

At present, because the statistical process monitoring methods based on principal component analysis do not need math model, it have been widely used in the field of process monitoring, but the ability of the fault diagnosis based on principal component analysis is limited. Aiming at this problem, fault reconstruction technique was used to reconsitute the T^2 statistics based on principal component analysis from the perspective of fault subspace and principal component analysis monitoring mode in this paper. The fault reconfigurable theory condition of the T2 statistics in the principal component subspace was obtained, and the specific fault identification index and diagnosis algorithm were proposed. The problem of fault diagnosis based on principal component subspace fault reconstruction technique had been solved, and it made up for the method of Dania et al. who had discussed the problem of fault reconstruction and recognition only in the residual subspace. Through the simulation monitoring of double-effect evaporator process, these results showed that the acquired theory condition, fault identification index and diagnosis algorithm could effectively identify sensor and process fault, and indicate the acquired theory, identification index and diagnosis algorithm were effective.