Date of Award
Level of Access Assigned by Author
Master of Science (MS)
Mohamad T. Musavi
Second Committee Member
Third Committee Member
The central idea of Principal Component Analysis (PCA) is to reduce the dimensionality of a data set through a linear transformation. PCA has been commonly used for process monitoring by extracting linear features (principal components) of a normal operation and using the resulting features to detect abnormal process deviations. However, conventional (linear) PCA fails to provide principal components that adequately represent nonlinear relationships in a system. In this thesis, a non-linear PCA (NLPCA) method is developed using artificial neural networks to address the limitations of linear PCA. The performance of PCA and NLPCA is compared in process monitoring applications using synthetic and real-world data.
Shannak, Kamal Majed, "On Non-Linear Principal Component Analysis for Process Monitoring" (2004). Electronic Theses and Dissertations. 906.