Análise comparativa entre RNA, AG e Migha na determinação de rugosidades através de calibração de redes hidráulicas

Abstract

This paper performs a comparative analysis between three methods to estimate roughness of water supply networks by calibration of hydraulics networks. In the first method, the artificial neural network (ANN) optimization tool was adopted, while hydraulic balance is achieved through the dynamic library of EPANET. The second, known as iterative alternative hydraulic gradient (MIGHA), initially developed to calibrate hydrodynamic parameters was adapted in this work. The third tool is the widely used optimization GA. The correct procedure would be to acquire, for the field measurements, data from the temporal variation of hydraulic head and, use them to solve the inverse problem through a comparison between the observed and estimated hydraulic loads. The estimation methodology uses data acquired by means of a hypothetical network.