An Evaluation Method for Water Quality Based on the Improved SOM Neural Network

• Published in: Applied Mechanics and Materials Vol. 373-375; pp. 1220 - 1223
• Main Authors: Li, Ze Xi, Wang, Chao Jie, Han, Yuan Feng, He, Jing Cheng, Li, Hong Yi, Zhao, Di
• Format: Journal Article
• Language: English
• Published: Zurich Trans Tech Publications Ltd 01.08.2013
• Subjects: SOM Neural Network; Correlation Matrix; Water Quality Evaluation; Principal Component Analysis (PCA)
• ISBN: 3037858060; 9783037858066
• ISSN: 1660-9336; 1662-7482; 1662-7482
• DOI: 10.4028/www.scientific.net/AMM.373-375.1220

In recent years, people have been paying increasingly attention on monitoring the quality of drinking water, which becomes rather necessary after natural disasters such as the Beijing 7.21 rainstorm, considering that the drinking water is one of the main medium for epidemic spreading. Most of the existing evaluation methods have their bases on concise mathematical models, which often fail to describe the complex essential nonlinear relations between the water quality and the chemical material in it. In this paper, we propose the evaluation method by using the SOM neural network, a unsupervised method that is able to classify, and therefore evaluate, given water samples. In order to promote the convergence rate and the precision of SOM neural network when dealing with high dimensional and highly correlated samples, we add a PCA preprocessing procedure. Experiment results demonstrate that the improved SOM neural network could evaluate the water quality with high precision.