Water quality failure is a long-standing problem worldwide, causing illness, poisoning, disease outbreak, and claiming human lives in the urban communities. Potable water can be compromised due to a myriad of physical, operational, and environmental factors, such as contaminants intrusion into water pipelines, leaching, disinfection byproducts, chemical or microbial permeation, and pollution. The prediction of potable water quality has seldom been researched, while a novel automated model that offers a proactive approach can be developed to promote sustainability-based strategies. This paper elaborates the impacts of the aforementioned factors on the quality of potable water using the Artificial Neural Networks (ANNs) and risk analysis techniques. The ANN model was developed based on historical data obtained from the water distribution networks (WDNs) of the City of El Pedregal, Peru. The data were streamlined and fed to the neural network to be trained. Subsequent to multiple iterations via the scaled conjugate gradient algorithm, the optimized performance was generated and passed to the trained network to forecast the water quality failure in WDNs. The model performance was tested and validated against different statistical error terms and indicators. The mean absolute error and root-mean-squared error in the ANN failure prediction were computed as 0.08 and 0.15, whereas the average validity of the network was generated to be 92%. Based on the trained neural network, the degree of influence of each factor was determined through implementing a sensitivity analysis. It was found that water quality, water pressure, and operational and maintenance practices had the maximum influence on the risk of failure in water infrastructure. Policy makers and managers can benefit from the proposed model since ANN is already trained, by predicting water quality whenever new data become available. Prediction results will indicate the level of risk (low, moderate, high) to the inhabitants, thereby preemptive measures can be taken to avoid any illness or disease outbreak.
Bibliographical noteFunding Information:
This work is supported by the collaboration of the Universidad Nacional de San Agust?n (UNSA) in Arequipa, Peru, Grant/Contract Number: 80522 and Purdue University in Indiana, USA through Discovery Park's Center for the Environment (C4E).
This work is supported by the collaboration of the Universidad Nacional de San Agustín (UNSA) in Arequipa, Peru , Grant/Contract Number: 80522 and Purdue University in Indiana, USA through Discovery Park’s Center for the Environment (C4E).
© 2020 Elsevier Ltd
- Artificial neural networks
- Clean water
- Water systems