FEATURES EXTRACTION FROM PRIMARY CLARIFIER USING UNSUPERVISED NEURAL NETWORKS (KOHONEN SELF ORGANISING MAP)

摘要

In this paper, the Kohonen Self-Organising Map (KSOM) was used to extract the most suitable features from a large array of data sampled for the primary clarifier of a wastewater treatment works in Edinburgh, UK. The extracted features were of a much smaller size than the original set, thus making it much easier to establish the correlation between the individual water quality variables. Such correlations will aid in designing intelligent models for the primary clarifier. The study demonstrates the efficiency of the KSOM as a tool for the discovery of correlation between large data sets, as well as the visualisation of such correlation.