생물환경연구실

논문 초록 (Abstract)

Multivariate statistical process control such as principal component ana lysis (PCA) and partial least squares (PLS) has been effectively utilize d to analyze large databases accumulated in industrial plants in order t o improve process performance and productquality. However, because both PCA and PLS are basically linear methods, nonlinearity in most chemical and biological processesis still a significant problem for their pract ical applications. Kernel-based algorithms are potentially very efficien t for monitoring process disturbances and predicting key quality variabl es of nonlinear processes by mapping an original input space into ahigh -dimensional feature space. Nonlinear data structure in the original spa ce is most likely to be linear at the high-dimensional feature space. Ke rnel-based methods essentially require only linear algebra, making them as simple as linear multivariate projection methods, and can handle a w ide range of nonlinearities because of their ability to use different ke rnel functions. In this work, kernel-based algorithms such as kernel PCA (KPCA) and kernel PLS (KPLS) are applied to a full-scale biological ana erobic filter process treating high-strength organic wastewater in a pet rochemical plant. KPCA is applied to detect process disturbances in real -time and KPLS to predict inferentially key process variables in the ana erobic filter process. The proposed kernel-based approaches could effect ively capture the nonlinear relationship in the process variables and sh ow far better performance in process monitoring and prediction of qualit y variables compared to the conventional PCA and PLS methods. ？ 2006 Am erican Chemical Society.