Cluster analysis by self-organizing maps: An application to the modelling of water quality in a treatment process

Abstract

The unit processes in water treatment involve many complex physical and chemical phenomena which are difficult to assess using traditional data analysis methods. Moreover, measurement data gathered from the process is often challenging with respect to modelling purposes, because there is a lack of continuous online measurements, for which sparse laboratory measurement data have to be conducted to compensate them.This paper reports on the application of self-organizing map (SOM) techniques combined with K-means clustering to model water quality in the treatment of drinking water. At the first phase of the study, a SOM was produced by using both on-line and laboratory data of the treatment process and raw water. At the second phase, the reference vectors of the map were classified by K-means algorithm into clusters, which can be used to present different states of the process. At the final phase, the results were interpreted by analyzing the reference vectors in the clusters. The introduced approach offers a straightforward method for assessing the essential characteristics of the process. In addition, the results clearly demonstrate some challenges in the modelling of water quality in treatment processes.