| Summary: | 碩士 === 國立臺灣大學 === 化學工程學研究所 === 88 === In the field of process monitoring, we often combine the statistical process control tools and the multivariate analysis techniques together to get better monitoring performance. There are many researches from which a typical pattern of process monitoring has been established.
There are several useful multivariate analysis techniques like principal component analysis (PCA), principal component regression (PCR), partial least square or projection to latent structure (PLS), canonical variable analysis (CVA) and etc. The aim of these techniques is to reduce the complexity of the multivariate process monitoring problem while establishing the process model.
In this artical, we propose a new monitoring method named transformed principal component analysis (TPCA), which was designed to compensate the effect of the existing of process delay. Then, we apply it to a illustrating example to test the fault detection performance of this method. Finally, we summarize the advantages and the trade-off of this method at the end of this artical as the conclusion.
|
|---|
