The effect of insufficient excitation in PCA estimation

Proceedings of the American Control Conference - Tập 4 - Trang 2707-2712 vol.4 - 2002
Jin Cao1, J. Gertler1
1Electrical and Computer Engineering Department, George Mason University, Fairfax VA, USA

Tóm tắt

Linear relations external to the monitored plant, such as ratio-control, feedback or upstream constraints, make the data non-persistently exciting for the identification of the plant model. The effect of such external relations is studied on the PCA estimation of linear static models, and compared to LS estimation. Also, ways to detect the presence of such relations in the PCA framework are suggested. The results are extended to persistency and order/degree estimation in linear dynamic and polynomial static models.

Từ khóa

#Principal component analysis #Input variables #Fault diagnosis #Least squares methods #Polynomials #Actuators #Data engineering #Computerized monitoring #Fault detection #Gold

Tài liệu tham khảo

piovoso, 1992, Monitoring process performance in real time, Amer Control Conf Chicago, 2359 10.1016/0959-1524(96)00009-1 ljung, 1987, System Identification Theory for the User macgregor, 1989, Multivariate statistical methods for monitoring large datasets from chemical processes, AIChE Meeting San Francisco 10.1016/S0959-1524(00)00041-X ljung, 1999, System Identification Theory for the User 2nd Edition 10.1016/0169-7439(95)00076-3 1995, Chemometrics and Intelligent Laboratory Systems, 30, 179, 10.1016/0169-7439(95)00076-3 jollife, 1986, Principal Component Analysis, 10.1007/978-1-4757-1904-8 10.1002/cjce.5450690105 10.1016/S0959-1524(02)00016-1 fang, 1988, Process Identification 10.2307/2287575 cao, 2001, Noise-induced bias in PCA modeling of linear system, American Control Conference Arlington VA USA 10.1109/TAC.1974.1100705 huang, 1999, Fault isolation by partial PCA and NLPCA, IFAC World Congress Beijing, o, 545 gertler, 2000, Structured partial principal component analysis: Extension to polynomial nonlinearities, IFAC Safeprocess Symposium Budapest Hungary 10.1002/aic.690450213 gertler, 1998, Fault Detection and Diagnosis in Engineering Systems gertler, 1997, Principal component analysis and parity relations-a strong duality, IFAC SAFEPROCESS Hull UK, 2, 837 10.1016/S0959-1524(99)00062-1 gilbert, 1995, Linear Algebra and Matrix Theory