Reliability modelling and analysis of water distribution network based on backpropagation recursive processes with real field data

Abstract

Any water distribution network (WDN) is a key element in critical infrastructure. As time series data mining and modelling are developing apace, these provide promising tools for the research of water resources management. The aim of this article is to explain how incomplete information on the failures in water mains can be used and processed effectively. The data available and under investigation are the truncated failure field data of a regional water distribution network recorded during the last 15 years. We introduce the application of novel, non-trivial dynamic backpropagation recursive time-series models which are later successfully validated by the recorded data. Real field data of a WDN were used while these data were elaborated with a backpropagation Kalman recursor. The results can be used for predicting the reliability of a WDN, or as inputs into a water management system to optimise maintenance, crisis management and emergency planning.