Influence of scaling and unfolding in PCA based monitoring of nutrient removing batch process

The dataset of batch biological and biotechnological processes can be organized in a three-way data matrix. In this chapter, the usefulness of different principal component analysis (PCA) approaches for monitoring is analyzed. Different ways of unfolding and scaling of data are applied to a pilot-scale sequencing batch reactor (SBR) data. PCA is used to reduce the dimensionality and to remove the nonlinearity dynamic of the data. Also, a new method to select the number of principal components is proposed. Loadings graphics are used to determinate the predominant variables for each one. The results show that whatever model can be applied depending on the goal of the monitoring, however, the models implicate possible false alarms or faults omission. © 2007