Comparison of two water quality modeling approaches in the trunk network of Bogota's water distribution system: white-box models versus black-box models using artificial intelligence

Abstract

La modelación de cloro y trihalometanos en redes de distribución de agua potable ha sido abordada principalmente mediante modelos de caja blanca. En este sentido, esta tesis presenta modelos de caja negra basados en inteligencia artificial (ANNs y ANFIS) como un medio para superar la complejidad de calibrar modelos de caja blanca al predecir estas variables a lo largo de toda la red teniendo la información de algunos puntos como variables de entrada.

The chlorine and TTHMs modeling in water distribution networks has been mainly addressed through white-box models. In this sense, this thesis presents black-box models based on artificial intelligence (ANNs and ANFIS) as a means to overcome the complexity of calibrating white-box models by predicting these variables throughout the network having the information of a few points as an input.