Self-organizing map and clustering for wastewater treatment monitoring

The objective of this project is the development of plant supervision techniques based on self-organizing map (SOM) for the implementation in a wastewater treatment plant. SOM is an unsupervised learning algorithm to establish the relationships among process variables. Clustering techniques such as K-means algorithm have been used for the system state estimation, monitoring and visualization of process states. The best clustering structure is selected by means of the Davies-Bouldin index for evaluation of the several structures obtained from K-means. (C) 2004 Elsevier Ltd. All rights reserved.