Using SOM and PCA for analysing and interpreting data from a P-removal SBR

This paper focuses on the application of Kohonen self-organizing maps (SOM) and principal component analysis (PCA) to thoroughly analyse and interpret multidimensional data from a biological process. The process is aimed at enhanced biological phosphorus removal (EBPR) from wastewater. In this work, SOM and PCA are firstly applied to the data set in order to identify and analyse the relationships among the variables in the process. Afterwards, K-means algorithm is used to find out how the observations can be grouped, on the basis of their similarity, in different classes. Finally, the information obtained using these intelligent tools is used for process interpretation and diagnosis. In the data set analysed, both techniques yielded similar results regarding the relationships among the variables and the clustering of the observations (i.e., the same groups of observations were identified) and, therefore, identical process interpretation could be made. The cluster analysis allowed relating the observations to process behaviour, clearly distinguishing start-up, desirable and poor process conditions. The results demonstrate that the applied techniques are highly effective to compress multidimensional data sets and to extract relevant information from the process, making the interpretation and diagnosis much easier and evident.